EDA¶

Jessica Stoker¶

Table of contents¶

  • Introduction

  • Setup

  • Data Manipulation

    • Combining Dataframes
    • Setting up for Comparison
    • Prices Combined Summaries
    • Percent Change Summaries

  • Plots

    • Individual Plots
    • Grouped Plots
    • Seasonality and Trends

  • Results of EDA

  • Modeling

    • ARIMA
    • ARIMA Results Comments
    • LSTM
    • Final Results

Business Problem ¶

Fluctuating prices in Swire’s supply chain results in lower margins. Building accurate forecasts of commodities used in production will help with procurement, production planning, and inventory management. Nate, Robyn, and I have been tasked to produce a solution to this problem. We aim to use time series forecasting methods to produce both short (3 months) and long-term (1 year) forecasts on a dashboard that can simulate profitability scenarios. Success will be measured by the ability of key stakeholders to use the dashboard and make informed decisions in the productions process, as well as the accuracy of the forecasts in the model.

In this stage of our project, we have downloaded as much historical data as possible from the sources provided. Our goal is to explore these data in hopes of getting a clearer picture of the previous fluctuations in the commodities prices. This will allow us to better create the models that will predict the future prices of these commodities.

To begin, I will begin by listing the questions that I am hoping to answer during my exploration.

  1. What do these data look like?
  2. What are the trends for each commodity?
  3. What are the monthly averages of each commodity?
  4. What about yearly averages?
  5. Are there seasonal trends for these commodities?
  6. Can I compare the fluctuation in these commodities?

Setup ¶

In [5]:
from google.colab import drive
drive.mount('/content/drive')

Drive already mounted at /content/drive; to attempt to forcibly remount, call drive.mount("/content/drive", force_remount=True).
In [6]:
# supress the display of warning messages
import warnings
warnings.filterwarnings('ignore')
In [7]:
pip install pmdarima

Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/
Collecting pmdarima
  Downloading pmdarima-2.0.3-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.manylinux_2_28_x86_64.whl (1.9 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.9/1.9 MB 24.4 MB/s eta 0:00:00
Requirement already satisfied: joblib>=0.11 in /usr/local/lib/python3.9/dist-packages (from pmdarima) (1.2.0)
Requirement already satisfied: scikit-learn>=0.22 in /usr/local/lib/python3.9/dist-packages (from pmdarima) (1.2.2)
Requirement already satisfied: Cython!=0.29.18,!=0.29.31,>=0.29 in /usr/local/lib/python3.9/dist-packages (from pmdarima) (0.29.34)
Requirement already satisfied: pandas>=0.19 in /usr/local/lib/python3.9/dist-packages (from pmdarima) (1.5.3)
Requirement already satisfied: numpy>=1.21.2 in /usr/local/lib/python3.9/dist-packages (from pmdarima) (1.24.2)
Requirement already satisfied: urllib3 in /usr/local/lib/python3.9/dist-packages (from pmdarima) (1.26.15)
Requirement already satisfied: setuptools!=50.0.0,>=38.6.0 in /usr/local/lib/python3.9/dist-packages (from pmdarima) (67.6.1)
Requirement already satisfied: statsmodels>=0.13.2 in /usr/local/lib/python3.9/dist-packages (from pmdarima) (0.13.5)
Requirement already satisfied: scipy>=1.3.2 in /usr/local/lib/python3.9/dist-packages (from pmdarima) (1.10.1)
Requirement already satisfied: python-dateutil>=2.8.1 in /usr/local/lib/python3.9/dist-packages (from pandas>=0.19->pmdarima) (2.8.2)
Requirement already satisfied: pytz>=2020.1 in /usr/local/lib/python3.9/dist-packages (from pandas>=0.19->pmdarima) (2022.7.1)
Requirement already satisfied: threadpoolctl>=2.0.0 in /usr/local/lib/python3.9/dist-packages (from scikit-learn>=0.22->pmdarima) (3.1.0)
Requirement already satisfied: patsy>=0.5.2 in /usr/local/lib/python3.9/dist-packages (from statsmodels>=0.13.2->pmdarima) (0.5.3)
Requirement already satisfied: packaging>=21.3 in /usr/local/lib/python3.9/dist-packages (from statsmodels>=0.13.2->pmdarima) (23.1)
Requirement already satisfied: six in /usr/local/lib/python3.9/dist-packages (from patsy>=0.5.2->statsmodels>=0.13.2->pmdarima) (1.16.0)
Installing collected packages: pmdarima
Successfully installed pmdarima-2.0.3
In [8]:
import pandas as pd
import numpy as np
import matplotlib
import matplotlib.pyplot as plt
plt.style.use('ggplot')
import seaborn as sns
from datetime import datetime
from statsmodels.tsa.seasonal import seasonal_decompose
from random import randrange
from pandas import Series
from statsmodels.graphics.tsaplots import plot_acf, plot_pacf
from pmdarima import auto_arima
from statsmodels.tsa.arima.model import ARIMA
from sklearn.preprocessing import MinMaxScaler
from keras.models import Sequential
from keras.layers import LSTM
from keras.layers import Dense
from keras.layers import Dropout
from keras.models import load_model
from statsmodels.tsa.statespace.kalman_filter import PredictionResults
import math
from sklearn.metrics import mean_squared_error
from sklearn.metrics import r2_score
from sklearn.model_selection import GridSearchCV
from keras.wrappers.scikit_learn import KerasRegressor
In [9]:
np.random.seed(123456)
In [10]:
# Upload files into pandas dataframes
#Sugar
sugar = pd.read_csv("/content/drive/MyDrive/Colab Notebooks/EDA/sugar-prices-37yr.csv", index_col = 'date', parse_dates = True)
#Cotton
cotton = pd.read_csv("/content/drive/MyDrive/Colab Notebooks/EDA/cotton-prices-45yr.csv", index_col = 'date', parse_dates = True)
#Corn
corn = pd.read_csv("/content/drive/MyDrive/Colab Notebooks/EDA/corn-prices-59yr.csv", index_col = 'date', parse_dates = True)
#Coffee
coffee = pd.read_csv("/content/drive/MyDrive/Colab Notebooks/EDA/coffee-prices-45yr.csv", index_col = 'date', parse_dates = True)
#Soy Bean
soybean = pd.read_csv("/content/drive/MyDrive/Colab Notebooks/EDA/soybean-prices-45yr.csv", index_col = 'date', parse_dates = True)
#Soy Bean Oil
soybean_oil = pd.read_csv("/content/drive/MyDrive/Colab Notebooks/EDA/soybean-oil-prices-45yr.csv", index_col = 'date', parse_dates = True)
#Alumuinum
aluminum = pd.read_csv("/content/drive/MyDrive/Colab Notebooks/EDA/Aluminum-prices-20yr_rv.csv", index_col = 'Date', parse_dates = True)

Data Manipulation ¶

There are a couple of ways that I did manipulate the data before importing the data into my notebook.

  • There was a 0 value on 1/22/2007 in the Aluminum data set. I couldn't see a clear reason why this value would be 0. I assumed that this was an error. I used Excel to filter out all but the values from January of 2007 and excluded the missing value before taking the average of that month. I used that average to replace the 0 value.

  • I was having a hard time parsing the correct dates with the date format that was automatically extracted from each site. I change the date format in excel to YYYY-MM-DD so that all dates were parsed correctly at import.

Data Manipulation in Python¶

You can see below that there are dates that haven't happened yet included in the dataframes, these are all labeled as NA. I chose to ensure that there were no NA values that will cause me issues in plotting and viewing the data later.

In [11]:
cotton.isna()
Out[11]:
value
date
1972-08-22 False
1972-08-23 False
1972-08-24 False
1972-08-25 False
1972-08-28 False
... ...
2023-03-06 True
2023-03-07 True
2023-03-08 True
2023-03-09 True
2023-03-10 True

12718 rows × 1 columns

In [12]:
#Drop NA rows
sugar = sugar.dropna()
cotton = cotton.dropna()
corn = corn.dropna()
cofee = coffee.dropna()
soybean = soybean.dropna()
soybean_oil = soybean_oil.dropna()
aluminum = aluminum.dropna()
#Rename value column to the comodity name for identification in combned dataframe
sugar = sugar.rename(columns={' value': 'sugar'})
cotton = cotton.rename(columns={' value': 'cotton'})
corn = corn.rename(columns={' value': 'corn'})
coffee = coffee.rename(columns={' value': 'coffee'})
soybean = soybean.rename(columns={' value': 'soybean'})
soybean_oil = soybean_oil.rename(columns={' value': 'soybean_oil'})
aluminum = aluminum.rename(columns={'Close': 'aluminum'})

Combining dataframes ¶

I chose to combine these dataframes to reduce the amount of code to write and for easier comparison. You will see in the next section that this merge only combines the rows that are represented in all dataframes.

In [13]:
#Merge all dataframes into one
all_prices = pd.merge(sugar, cotton, left_index=True, right_index=True).merge(corn, left_index=True, right_index=True).merge(coffee,left_index=True, right_index=True).merge(soybean,left_index=True, right_index=True).merge(soybean_oil,left_index=True, right_index=True).merge(aluminum, left_index=True, right_index=True)
In [14]:
#Creating dataframe with monthly average
all_prices_M = all_prices.resample('M').mean()
#Creating dataframe with yearly average
all_prices_Y = all_prices.resample('Y').mean()
In [15]:
#Correlation heatmap
corr = all_prices.corr()

# Create heatmap
sns.heatmap(corr, annot=True, cmap='Blues')
plt.title(f'Correlation Heatmap')
Out[15]:
Text(0.5, 1.0, 'Correlation Heatmap')

Setting up for comparison ¶

I wanted to see how the price fluctuations compare to each other. Because the prices are all very different, I chose percentage of change to compare the fluctuations in price across all commodities.

In [16]:
#Percent change of all comoditiees for all three comodities for later comparison
all_prices_percent_change = all_prices.pct_change()
all_prices_percent_change_M = all_prices_M.pct_change()
all_prices_percent_change_Y = all_prices_Y.pct_change()

Prices combined Summaries ¶

In [17]:
#Daily
all_prices.head()
Out[17]:
sugar cotton corn coffee soybean soybean_oil aluminum
2002-10-23 0.0710 0.4521 2.5025 0.6430 5.5450 0.2084 1326.0
2002-10-25 0.0699 0.4663 2.5025 0.6855 5.5700 0.2119 1338.0
2002-10-28 0.0713 0.4631 2.4750 0.6470 5.4565 0.2087 1348.0
2002-10-29 0.0727 0.4622 2.4900 0.6290 5.5790 0.2181 1333.0
2002-10-30 0.0718 0.4610 2.4600 0.6300 5.5775 0.2168 1336.0
In [18]:
all_prices.tail()
Out[18]:
sugar cotton corn coffee soybean soybean_oil aluminum
2022-06-24 0.1837 1.0376 7.5025 2.2660 16.1075 0.6975 2455.85
2022-06-27 0.1829 0.9994 7.4970 2.2680 16.0849 0.7023 2474.77
2022-06-28 0.1826 1.0196 7.4406 2.2478 16.3152 0.7085 2490.85
2022-06-29 0.1848 1.0295 7.5882 2.2279 16.6662 0.7074 2470.85
2022-06-30 0.1855 1.0492 7.6969 2.3280 16.7046 0.7012 2431.33
In [19]:
all_prices.describe()
Out[19]:
sugar cotton corn coffee soybean soybean_oil aluminum
count 4681.000000 4681.000000 4681.000000 4681.000000 4681.000000 4681.000000 4681.000000
mean 0.153625 0.741851 4.238816 1.369563 10.328070 0.379300 2060.932749
std 0.054323 0.249296 1.518258 0.459407 2.948649 0.128631 442.617220
min 0.054000 0.391400 1.907500 0.560500 5.012500 0.189400 1288.000000
25% 0.116300 0.587800 3.455000 1.094500 8.637500 0.288300 1763.850000
50% 0.146100 0.690600 3.780000 1.257100 9.835000 0.336900 1951.000000
75% 0.184600 0.827100 5.235000 1.523500 12.825000 0.475500 2365.000000
max 0.353100 2.141400 8.312500 3.061500 17.690000 0.882000 3849.000000
In [20]:
#Monthly
all_prices_M.head()
Out[20]:
sugar cotton corn coffee soybean soybean_oil aluminum
2002-10-31 0.071717 0.461200 2.484167 0.649000 5.565083 0.213717 1338.166667
2002-11-30 0.073406 0.481375 2.427437 0.664056 5.697813 0.222244 1378.000000
2002-12-31 0.074810 0.503860 2.385750 0.629300 5.643600 0.216480 1376.800000
2003-01-31 0.078794 0.510139 2.352083 0.649583 5.641944 0.205850 1379.833333
2003-02-28 0.087317 0.517767 2.363778 0.628133 5.701806 0.204706 1409.333333
In [21]:
all_prices_M.tail()
Out[21]:
sugar cotton corn coffee soybean soybean_oil aluminum
2022-02-28 0.181650 1.239125 6.500600 2.460945 15.885375 0.670290 3213.762500
2022-03-31 0.191078 1.253026 7.465652 2.228587 16.733152 0.740065 3509.552174
2022-04-30 0.196332 1.395416 7.842658 2.256816 16.778526 0.787295 3266.242105
2022-05-31 0.192860 1.449710 7.828875 2.181675 16.590250 0.810505 2860.063500
2022-06-30 0.187100 1.321185 7.629135 2.304260 16.850170 0.754965 2582.123500
In [22]:
all_prices_M.describe()
Out[22]:
sugar cotton corn coffee soybean soybean_oil aluminum
count 237.000000 237.000000 237.000000 237.000000 237.000000 237.000000 237.000000
mean 0.150682 0.736104 4.175614 1.346301 10.204920 0.374718 2044.782723
std 0.055502 0.247763 1.524504 0.465912 2.972430 0.128677 442.351181
min 0.055880 0.423111 1.944275 0.590656 5.272344 0.197794 1338.166667
25% 0.115080 0.581089 3.321818 1.076543 8.562476 0.284026 1746.595238
50% 0.142877 0.691868 3.755684 1.248211 9.760955 0.332768 1933.388889
75% 0.182411 0.819568 5.153947 1.503632 12.815000 0.466889 2335.523684
max 0.320916 2.006878 8.038773 2.807953 16.850170 0.810505 3509.552174
In [23]:
#Yearly
all_prices_Y.head()
Out[23]:
sugar cotton corn coffee soybean soybean_oil aluminum
2002-12-31 0.073528 0.484619 2.425047 0.650372 5.655984 0.218844 1370.156250
2003-12-31 0.070143 0.598642 2.335328 0.624503 6.296221 0.226000 1425.805970
2004-12-31 0.075558 0.545429 2.496232 0.774228 7.267136 0.261391 1728.165563
2005-12-31 0.100474 0.507222 2.118782 1.079012 6.136122 0.227066 1902.215556
2006-12-31 0.146364 0.523831 2.625165 1.083871 5.945710 0.251239 2597.398041
In [24]:
all_prices_Y.tail()
Out[24]:
sugar cotton corn coffee soybean soybean_oil aluminum
2018-12-31 0.122719 0.820701 3.700973 1.136432 9.337805 0.299246 2113.245902
2019-12-31 0.123575 0.673787 3.857450 1.017841 8.932635 0.293129 1812.916452
2020-12-31 0.128944 0.643035 3.640972 1.111682 9.535121 0.312803 1731.591492
2021-12-31 0.178912 0.934919 5.742911 1.693772 13.684052 0.577515 2489.281653
2022-12-31 0.188943 1.307972 7.238638 2.298261 16.179177 0.729480 3082.201074
In [25]:
all_prices_Y.describe()
Out[25]:
sugar cotton corn coffee soybean soybean_oil aluminum
count 21.000000 21.000000 21.000000 21.000000 21.000000 21.000000 21.000000
mean 0.148914 0.739734 4.182742 1.344488 10.172620 0.377281 2045.969347
std 0.052455 0.236596 1.540538 0.478936 3.031101 0.136123 450.811774
min 0.070143 0.484619 2.118782 0.624503 5.655984 0.218844 1370.156250
25% 0.121284 0.575703 3.596669 1.083871 8.650238 0.293129 1728.165563
50% 0.146364 0.656322 3.757141 1.262322 9.785745 0.334154 1902.215556
75% 0.179519 0.820701 5.297314 1.636942 12.323684 0.458677 2420.197959
max 0.270685 1.356376 7.238638 2.527503 16.179177 0.729480 3082.201074

Percent Change Summaries ¶

In [26]:
#Daily
all_prices_percent_change.head()
Out[26]:
sugar cotton corn coffee soybean soybean_oil aluminum
2002-10-23 NaN NaN NaN NaN NaN NaN NaN
2002-10-25 -0.015493 0.031409 0.000000 0.066096 0.004509 0.016795 0.009050
2002-10-28 0.020029 -0.006863 -0.010989 -0.056163 -0.020377 -0.015101 0.007474
2002-10-29 0.019635 -0.001943 0.006061 -0.027821 0.022450 0.045041 -0.011128
2002-10-30 -0.012380 -0.002596 -0.012048 0.001590 -0.000269 -0.005961 0.002251
In [27]:
all_prices_percent_change.tail()
Out[27]:
sugar cotton corn coffee soybean soybean_oil aluminum
2022-06-24 -0.000544 -0.238850 0.004687 -0.025167 0.010984 0.030128 -0.008679
2022-06-27 -0.004355 -0.036816 -0.000733 0.000883 -0.001403 0.006882 0.007704
2022-06-28 -0.001640 0.020212 -0.007523 -0.008907 0.014318 0.008828 0.006498
2022-06-29 0.012048 0.009710 0.019837 -0.008853 0.021514 -0.001553 -0.008029
2022-06-30 0.003788 0.019136 0.014325 0.044930 0.002304 -0.008764 -0.015994
In [28]:
all_prices_percent_change.describe()
Out[28]:
sugar cotton corn coffee soybean soybean_oil aluminum
count 4680.000000 4680.000000 4680.000000 4680.000000 4680.000000 4680.000000 4680.000000
mean 0.000413 0.000348 0.000398 0.000488 0.000356 0.000381 0.000224
std 0.020373 0.018245 0.017797 0.020698 0.015475 0.015590 0.013698
min -0.116318 -0.238850 -0.078749 -0.106439 -0.099350 -0.088610 -0.080695
25% -0.011159 -0.008980 -0.009246 -0.011819 -0.007661 -0.008697 -0.007545
50% 0.000000 0.000170 0.000000 0.000363 0.000793 0.000000 0.000088
75% 0.011767 0.009943 0.010025 0.011791 0.008562 0.009539 0.008191
max 0.110230 0.123173 0.094257 0.138509 0.087185 0.084023 0.061368
In [29]:
#Monthly
all_prices_percent_change_M.head()
Out[29]:
sugar cotton corn coffee soybean soybean_oil aluminum
2002-10-31 NaN NaN NaN NaN NaN NaN NaN
2002-11-30 0.023559 0.043745 -0.022836 0.023199 0.023850 0.039899 0.029767
2002-12-31 0.019123 0.046710 -0.017173 -0.052339 -0.009515 -0.025934 -0.000871
2003-01-31 0.053261 0.012462 -0.014112 0.032232 -0.000293 -0.049104 0.002203
2003-02-28 0.108158 0.014952 0.004972 -0.033021 0.010610 -0.005560 0.021379
In [30]:
all_prices_percent_change_M.tail()
Out[30]:
sugar cotton corn coffee soybean soybean_oil aluminum
2022-02-28 -0.015251 0.035707 0.064914 0.038719 0.128265 0.100564 0.070333
2022-03-31 0.051903 0.011218 0.148456 -0.094418 0.053368 0.104097 0.092038
2022-04-30 0.027493 0.113637 0.050499 0.012667 0.002712 0.063818 -0.069328
2022-05-31 -0.017682 0.038909 -0.001757 -0.033295 -0.011221 0.029481 -0.124357
2022-06-30 -0.029866 -0.088656 -0.025513 0.056188 0.015667 -0.068525 -0.097180
In [31]:
all_prices_percent_change_M.describe()
Out[31]:
sugar cotton corn coffee soybean soybean_oil aluminum
count 236.000000 236.000000 236.000000 236.000000 236.000000 236.000000 236.000000
mean 0.007269 0.006867 0.007230 0.007682 0.006557 0.007021 0.004031
std 0.080502 0.068661 0.070095 0.069028 0.060832 0.057749 0.049564
min -0.277513 -0.300790 -0.244695 -0.165160 -0.214738 -0.222666 -0.193013
25% -0.042955 -0.033991 -0.028200 -0.039338 -0.026117 -0.029516 -0.026711
50% 0.004042 0.007121 0.007917 0.001954 0.002955 0.005719 0.007474
75% 0.052839 0.047522 0.047916 0.050387 0.038898 0.043294 0.035483
max 0.242327 0.211873 0.258090 0.311268 0.169568 0.152107 0.151857
In [32]:
#Yearly
all_prices_percent_change_Y.head()
Out[32]:
sugar cotton corn coffee soybean soybean_oil aluminum
2002-12-31 NaN NaN NaN NaN NaN NaN NaN
2003-12-31 -0.046041 0.235284 -0.036997 -0.039775 0.113196 0.032698 0.040616
2004-12-31 0.077197 -0.088889 0.068900 0.239749 0.154206 0.156602 0.212062
2005-12-31 0.329771 -0.070050 -0.151208 0.393663 -0.155634 -0.131319 0.100714
2006-12-31 0.456731 0.032745 0.238997 0.004503 -0.031031 0.106460 0.365459
In [33]:
all_prices_percent_change_Y.tail()
Out[33]:
sugar cotton corn coffee soybean soybean_oil aluminum
2018-12-31 -0.223043 0.120739 0.023329 -0.148687 -0.045775 -0.104468 0.067122
2019-12-31 0.006970 -0.179011 0.042280 -0.104354 -0.043390 -0.020440 -0.142118
2020-12-31 0.043447 -0.045639 -0.056119 0.092197 0.067448 0.067116 -0.044859
2021-12-31 0.387520 0.453915 0.577302 0.523611 0.435121 0.846256 0.437569
2022-12-31 0.056068 0.399022 0.260447 0.356890 0.182338 0.263136 0.238189
In [34]:
all_prices_percent_change_Y.describe()
Out[34]:
sugar cotton corn coffee soybean soybean_oil aluminum
count 20.000000 20.000000 20.000000 20.000000 20.000000 20.000000 20.000000
mean 0.076294 0.078802 0.084022 0.095009 0.070922 0.090559 0.058596
std 0.250358 0.252986 0.257902 0.261241 0.200115 0.269170 0.193479
min -0.323425 -0.410822 -0.292316 -0.305560 -0.231611 -0.305041 -0.349785
25% -0.146857 -0.089354 -0.064508 -0.064813 -0.047471 -0.109252 -0.053459
50% 0.049758 0.040866 0.022142 0.058539 0.037467 0.049907 0.033112
75% 0.258589 0.149375 0.244360 0.315569 0.161239 0.197779 0.216381
max 0.480152 0.632384 0.579747 0.544040 0.454870 0.846256 0.437569

Take aways for me from this section¶

  • It looks like the dataframes were properly merged; all columns are equal. They now start and end on the same date.
  • We can now see the basic summary statistics both in the original pricing values and the percentage of change.

Plots ¶

Individual Plots ¶

In [35]:
# Will print an inividual plot for all comodities by day
for col in all_prices.columns:
  all_prices[col].plot()
  plt.title(f'Price by Day {col}')
  plt.xlabel('Time - Day')
  plt.ylabel(f'{col}')
  plt.show()
In [36]:
# Will print an inividual plot for all comodities by month
for col in all_prices_M.columns:
  all_prices_M[col].plot()
  plt.title(f'Price by month {col}')
  plt.xlabel('Time - Month')
  plt.ylabel(f'{col} Price')
  plt.show()
In [37]:
# Will print an inividual plot for all comodities by Year
for col in all_prices_Y.columns:
  all_prices_Y[col].plot()
  plt.title(f'Price by year {col}')
  plt.xlabel('Time - Year')
  plt.ylabel(f'{col} Price')
  plt.show()

Takeaways from this section¶

You can see that there are moments in time that either have a direct or delayed reaction to major economic moments (~2008, ~2011, ~2020)

  • In 2011 cotton spiked due to extreme weather events in Pakistan(flood) and China(drought) as well as a restriction of exports in India.
  • Coffee was affected by extreme weather circumstances in 2022, 2011, 1986, and 1977.
  • Soybean oil and aluminum prices hit a historic high last year (2022). Soybeans also hit historic highs in 2021 and 2022.

Grouped Plots ¶

I wanted to see how the price flucuations compare to each other. Becuase the prices are all very different, put them on the same scale of measurement

In [38]:
#All dates
all_prices_percent_change.plot(figsize=(20,10))
plt.title(f'Percent Change by Day')
plt.show()

#By month
all_prices_percent_change_M.plot(figsize=(20,10))
plt.title(f'Percent Change by Month')
plt.show()
In [39]:
all_prices_percent_change_Y.plot(figsize=(20,10))
plt.title(f'Percent Change by Year')
plt.xlabel('Time - Year')
plt.ylabel('Percent Change')
plt.show()

Take aways from this section¶

  • I didn't learn a lot from the first two plots, there was too much overlapping information
  • The yearly plot in this case is the most interesting. To see that the prices do somewhat rise and fall together. This doesn't nessesarily mean that they covary. This could be explored purther with correlation plot

Seasonality and Trends ¶

In [40]:
for col in all_prices_M.columns:
  result = seasonal_decompose(all_prices_M[col], period=12)
  result.plot().set_size_inches(20, 10)
  plt.show()

Takeaways from this section¶

  • These charts are ment to give you a look into your data.
  • Will show any pattern of seasonality.
  • These plots are interesting in that it shows the differences in the trends of each commodity.

Modeling ¶

I would have liked to run 3 models: ARIMA, Long Short-Term Memory (LSTM), and GARCH models. These models all can make predictions for time series data. ARIMA isn't mentioned as the most effective for commodities, but it is the only model for time series forecasting that we learn in this program. I thought it would be a good comparison for more complicated models. The LSTM model is a deep learning neural network model, it is useful for long term trends in data. The GARCH model, which I did not manage to produce, is specifically applicable for forecasting commodities pricing.

ARIMA ¶

Preprocessing for ARIMA models¶

In [41]:
#Test and train split
x = len(all_prices_M)
size = int(x * 0.8)
index = x - size
all_prices_M_train = all_prices_M.iloc[:size,]
all_prices_M_test = all_prices_M.iloc[size:,]
In [42]:
#Checking the trian split
for col in all_prices_M_train.columns:
  col = pd.DataFrame(all_prices_M_train[col])
  print(col)
  print(len(col))

               sugar
2002-10-31  0.071717
2002-11-30  0.073406
2002-12-31  0.074810
2003-01-31  0.078794
2003-02-28  0.087317
...              ...
2018-02-28  0.135574
2018-03-31  0.128257
2018-04-30  0.117837
2018-05-31  0.118347
2018-06-30  0.120967

[189 rows x 1 columns]
189
              cotton
2002-10-31  0.461200
2002-11-30  0.481375
2002-12-31  0.503860
2003-01-31  0.510139
2003-02-28  0.517767
...              ...
2018-02-28  0.780116
2018-03-31  0.827152
2018-04-30  0.833284
2018-05-31  0.865279
2018-06-30  0.891467

[189 rows x 1 columns]
189
                corn
2002-10-31  2.484167
2002-11-30  2.427437
2002-12-31  2.385750
2003-01-31  2.352083
2003-02-28  2.363778
...              ...
2018-02-28  3.667842
2018-03-31  3.831429
2018-04-30  3.861605
2018-05-31  4.018816
2018-06-30  3.657452

[189 rows x 1 columns]
189
              coffee
2002-10-31  0.649000
2002-11-30  0.664056
2002-12-31  0.629300
2003-01-31  0.649583
2003-02-28  0.628133
...              ...
2018-02-28  1.209011
2018-03-31  1.197571
2018-04-30  1.173321
2018-05-31  1.203316
2018-06-30  1.161443

[189 rows x 1 columns]
189
              soybean
2002-10-31   5.565083
2002-11-30   5.697813
2002-12-31   5.643600
2003-01-31   5.641944
2003-02-28   5.701806
...               ...
2018-02-28  10.115000
2018-03-31  10.437262
2018-04-30  10.389105
2018-05-31  10.247368
2018-06-30   9.257214

[189 rows x 1 columns]
189
            soybean_oil
2002-10-31     0.213717
2002-11-30     0.222244
2002-12-31     0.216480
2003-01-31     0.205850
2003-02-28     0.204706
...                 ...
2018-02-28     0.322026
2018-03-31     0.319348
2018-04-30     0.312895
2018-05-31     0.311126
2018-06-30     0.298319

[189 rows x 1 columns]
189
               aluminum
2002-10-31  1338.166667
2002-11-30  1378.000000
2002-12-31  1376.800000
2003-01-31  1379.833333
2003-02-28  1409.333333
...                 ...
2018-02-28  2168.921053
2018-03-31  2089.166667
2018-04-30  2266.157895
2018-05-31  2296.842105
2018-06-30  2233.904762

[189 rows x 1 columns]
189
In [43]:
#Checking the test spilt
for col in all_prices_M_test.columns:
  col = pd.DataFrame(all_prices_M_test[col])
  print(col)
  print(len(col))

               sugar
2018-07-31  0.111552
2018-08-31  0.104595
2018-09-30  0.108732
2018-10-31  0.131848
2018-11-30  0.127924
2018-12-31  0.125621
2019-01-31  0.126810
2019-02-28  0.129316
2019-03-31  0.124738
2019-04-30  0.125820
2019-05-31  0.118224
2019-06-30  0.124675
2019-07-31  0.121286
2019-08-31  0.115614
2019-09-30  0.112480
2019-10-31  0.124574
2019-11-30  0.126870
2019-12-31  0.133300
2020-01-31  0.141667
2020-02-29  0.150421
2020-03-31  0.118050
2020-04-30  0.100550
2020-05-31  0.106739
2020-06-30  0.118368
2020-07-31  0.119018
2020-08-31  0.128220
2020-09-30  0.124838
2020-10-31  0.142877
2020-11-30  0.149335
2020-12-31  0.146529
2021-01-31  0.159163
2021-02-28  0.168879
2021-03-31  0.158057
2021-04-30  0.163395
2021-05-31  0.172200
2021-06-30  0.172427
2021-07-31  0.177148
2021-08-31  0.193438
2021-09-30  0.193445
2021-10-31  0.196190
2021-11-30  0.197462
2021-12-31  0.191786
2022-01-31  0.184463
2022-02-28  0.181650
2022-03-31  0.191078
2022-04-30  0.196332
2022-05-31  0.192860
2022-06-30  0.187100
48
              cotton
2018-07-31  0.875357
2018-08-31  0.845933
2018-09-30  0.806258
2018-10-31  0.774678
2018-11-30  0.772752
2018-12-31  0.773389
2019-01-31  0.730781
2019-02-28  0.716868
2019-03-31  0.753429
2019-04-30  0.773385
2019-05-31  0.693886
2019-06-30  0.656425
2019-07-31  0.636868
2019-08-31  0.590348
2019-09-30  0.597740
2019-10-31  0.634757
2019-11-30  0.641115
2019-12-31  0.667875
2020-01-31  0.700610
2020-02-29  0.671037
2020-03-31  0.575464
2020-04-30  0.537460
2020-05-31  0.572478
2020-06-30  0.608200
2020-07-31  0.626905
2020-08-31  0.635080
2020-09-30  0.643267
2020-10-31  0.691868
2020-11-30  0.700885
2020-12-31  0.746429
2021-01-31  0.808489
2021-02-28  0.866105
2021-03-31  0.852774
2021-04-30  0.842821
2021-05-31  0.848895
2021-06-30  0.854636
2021-07-31  0.891333
2021-08-31  0.935510
2021-09-30  0.955909
2021-10-31  1.086890
2021-11-30  1.174419
2021-12-31  1.076333
2022-01-31  1.196405
2022-02-28  1.239125
2022-03-31  1.253026
2022-04-30  1.395416
2022-05-31  1.449710
2022-06-30  1.321185
48
                corn
2018-07-31  3.530238
2018-08-31  3.609810
2018-09-30  3.584474
2018-10-31  3.681522
2018-11-30  3.669548
2018-12-31  3.815789
2019-01-31  3.786905
2019-02-28  3.747789
2019-03-31  3.711071
2019-04-30  3.588100
2019-05-31  3.845952
2019-06-30  4.357575
2019-07-31  4.280114
2019-08-31  3.781048
2019-09-30  3.678250
2019-10-31  3.897717
2019-11-30  3.746575
2019-12-31  3.829750
2020-01-31  3.857381
2020-02-29  3.777579
2020-03-31  3.588636
2020-04-30  3.212775
2020-05-31  3.196389
2020-06-30  3.278614
2020-07-31  3.311591
2020-08-31  3.248075
2020-09-30  3.662738
2020-10-31  3.988977
2020-11-30  4.163025
2020-12-31  4.360000
2021-01-31  5.153947
2021-02-28  5.494184
2021-03-31  5.465000
2021-04-30  6.164737
2021-05-31  6.757763
2021-06-30  6.612318
2021-07-31  5.573214
2021-08-31  5.527071
2021-09-30  5.243864
2021-10-31  5.366190
2021-11-30  5.710976
2021-12-31  5.911548
2022-01-31  6.104342
2022-02-28  6.500600
2022-03-31  7.465652
2022-04-30  7.842658
2022-05-31  7.828875
2022-06-30  7.629135
48
              coffee
2018-07-31  1.110571
2018-08-31  1.033967
2018-09-30  0.999289
2018-10-31  1.148500
2018-11-30  1.121786
2018-12-31  1.029947
2019-01-31  1.033500
2019-02-28  0.998205
2019-03-31  0.964738
2019-04-30  0.920395
2019-05-31  0.927524
2019-06-30  1.008205
2019-07-31  1.057864
2019-08-31  0.949200
2019-09-30  0.998975
2019-10-31  0.973674
2019-11-30  1.095995
2019-12-31  1.297275
2020-01-31  1.135690
2020-02-29  1.039121
2020-03-31  1.149773
2020-04-30  1.134695
2020-05-31  1.058250
2020-06-30  0.968100
2020-07-31  1.039977
2020-08-31  1.183635
2020-09-30  1.208190
2020-10-31  1.074023
2020-11-30  1.120555
2020-12-31  1.228810
2021-01-31  1.244000
2021-02-28  1.274137
2021-03-31  1.300326
2021-04-30  1.334811
2021-05-31  1.503632
2021-06-30  1.563345
2021-07-31  1.695429
2021-08-31  1.813529
2021-09-30  1.902364
2021-10-31  2.027690
2021-11-30  2.213105
2021-12-31  2.351548
2022-01-31  2.369211
2022-02-28  2.460945
2022-03-31  2.228587
2022-04-30  2.256816
2022-05-31  2.181675
2022-06-30  2.304260
48
              soybean
2018-07-31   8.528381
2018-08-31   8.684024
2018-09-30   8.393947
2018-10-31   8.607457
2018-11-30   8.828571
2018-12-31   9.021868
2019-01-31   9.135595
2019-02-28   9.117421
2019-03-31   9.017024
2019-04-30   8.840575
2019-05-31   8.369762
2019-06-30   8.894325
2019-07-31   8.878045
2019-08-31   8.600476
2019-09-30   8.832750
2019-10-31   9.262891
2019-11-30   9.124250
2019-12-31   9.114050
2020-01-31   9.221667
2020-02-29   8.866658
2020-03-31   8.721477
2020-04-30   8.437050
2020-05-31   8.429722
2020-06-30   8.667523
2020-07-31   8.938341
2020-08-31   8.984050
2020-09-30   9.962976
2020-10-31  10.546659
2020-11-30  11.447500
2020-12-31  12.049095
2021-01-31  13.707763
2021-02-28  13.821263
2021-03-31  14.137283
2021-04-30  14.672000
2021-05-31  15.589342
2021-06-30  14.605568
2021-07-31  14.105238
2021-08-31  13.401714
2021-09-30  12.815000
2021-10-31  12.313667
2021-11-30  12.428333
2021-12-31  12.856667
2022-01-31  14.079474
2022-02-28  15.885375
2022-03-31  16.733152
2022-04-30  16.778526
2022-05-31  16.590250
2022-06-30  16.850170
48
            soybean_oil
2018-07-31     0.283748
2018-08-31     0.283443
2018-09-30     0.280289
2018-10-31     0.290517
2018-11-30     0.277138
2018-12-31     0.284026
2019-01-31     0.290010
2019-02-28     0.302574
2019-03-31     0.293805
2019-04-30     0.285805
2019-05-31     0.272352
2019-06-30     0.278660
2019-07-31     0.280673
2019-08-31     0.285852
2019-09-30     0.290775
2019-10-31     0.302800
2019-11-30     0.309735
2019-12-31     0.326305
2020-01-31     0.331510
2020-02-29     0.302879
2020-03-31     0.270177
2020-04-30     0.262090
2020-05-31     0.267494
2020-06-30     0.279073
2020-07-31     0.291905
2020-08-31     0.315160
2020-09-30     0.334929
2020-10-31     0.332768
2020-11-30     0.368290
2020-12-31     0.393967
2021-01-31     0.431542
2021-02-28     0.468142
2021-03-31     0.533700
2021-04-30     0.581337
2021-05-31     0.657753
2021-06-30     0.650786
2021-07-31     0.648138
2021-08-31     0.614771
2021-09-30     0.572768
2021-10-31     0.612462
2021-11-30     0.592338
2021-12-31     0.551038
2022-01-31     0.609042
2022-02-28     0.670290
2022-03-31     0.740065
2022-04-30     0.787295
2022-05-31     0.810505
2022-06-30     0.754965
48
               aluminum
2018-07-31  2064.500000
2018-08-31  2072.119048
2018-09-30  2056.052632
2018-10-31  2040.826087
2018-11-30  1951.023810
2018-12-31  1922.763158
2019-01-31  1868.119048
2019-02-28  1889.894737
2019-03-31  1895.095238
2019-04-30  1864.975000
2019-05-31  1809.547619
2019-06-30  1782.875000
2019-07-31  1819.954545
2019-08-31  1770.190476
2019-09-30  1780.925000
2019-10-31  1733.441304
2019-11-30  1768.544000
2019-12-31  1781.937500
2020-01-31  1791.771429
2020-02-29  1712.788947
2020-03-31  1633.948636
2020-04-30  1496.876500
2020-05-31  1498.403889
2020-06-30  1589.969545
2020-07-31  1679.947727
2020-08-31  1774.656500
2020-09-30  1780.501429
2020-10-31  1822.025455
2020-11-30  1945.948000
2020-12-31  2027.784286
2021-01-31  2005.817368
2021-02-28  2083.313158
2021-03-31  2213.614348
2021-04-30  2335.523684
2021-05-31  2454.326842
2021-06-30  2448.161818
2021-07-31  2507.321429
2021-08-31  2596.131905
2021-09-30  2850.270909
2021-10-31  2949.940952
2021-11-30  2635.745714
2021-12-31  2699.553333
2022-01-31  3002.581053
2022-02-28  3213.762500
2022-03-31  3509.552174
2022-04-30  3266.242105
2022-05-31  2860.063500
2022-06-30  2582.123500
48
In [44]:
#Auto Correlation and Partial autocorrelation charts for exploration
for col in all_prices_M_train.columns:
  print(col)
  plot_acf(all_prices_M_train[col], lags = 50)
  plt.ylim(-1, 1.2)
  plt.show()
  plot_pacf(all_prices_M_train[col], lags = 50)
  plt.ylim(-1, 1.1)
  plt.show()

sugar

cotton

corn

coffee

soybean

soybean_oil

aluminum
In [45]:
for col in all_prices_M_train.columns:
  print(col)
  auto_arima(all_prices_M_train[col], trace = True)

sugar
Performing stepwise search to minimize aic
 ARIMA(2,1,2)(0,0,0)[0] intercept   : AIC=-1091.029, Time=0.42 sec
 ARIMA(0,1,0)(0,0,0)[0] intercept   : AIC=-1056.831, Time=0.04 sec
 ARIMA(1,1,0)(0,0,0)[0] intercept   : AIC=-1083.770, Time=0.12 sec
 ARIMA(0,1,1)(0,0,0)[0] intercept   : AIC=-1096.112, Time=0.18 sec
 ARIMA(0,1,0)(0,0,0)[0]             : AIC=-1058.769, Time=0.03 sec
 ARIMA(1,1,1)(0,0,0)[0] intercept   : AIC=-1095.024, Time=0.18 sec
 ARIMA(0,1,2)(0,0,0)[0] intercept   : AIC=-1094.932, Time=0.13 sec
 ARIMA(1,1,2)(0,0,0)[0] intercept   : AIC=-1092.904, Time=0.23 sec
 ARIMA(0,1,1)(0,0,0)[0]             : AIC=-1098.079, Time=0.08 sec
 ARIMA(1,1,1)(0,0,0)[0]             : AIC=-1096.269, Time=0.04 sec
 ARIMA(0,1,2)(0,0,0)[0]             : AIC=-1096.878, Time=0.04 sec
 ARIMA(1,1,0)(0,0,0)[0]             : AIC=-1085.738, Time=0.02 sec
 ARIMA(1,1,2)(0,0,0)[0]             : AIC=-1094.866, Time=0.11 sec

Best model:  ARIMA(0,1,1)(0,0,0)[0]          
Total fit time: 1.641 seconds
cotton
Performing stepwise search to minimize aic
 ARIMA(2,1,2)(0,0,0)[0] intercept   : AIC=-487.143, Time=0.37 sec
 ARIMA(0,1,0)(0,0,0)[0] intercept   : AIC=-470.801, Time=0.06 sec
 ARIMA(1,1,0)(0,0,0)[0] intercept   : AIC=-483.663, Time=0.06 sec
 ARIMA(0,1,1)(0,0,0)[0] intercept   : AIC=-479.722, Time=0.07 sec
 ARIMA(0,1,0)(0,0,0)[0]             : AIC=-472.591, Time=0.03 sec
 ARIMA(1,1,2)(0,0,0)[0] intercept   : AIC=-484.790, Time=0.26 sec
 ARIMA(2,1,1)(0,0,0)[0] intercept   : AIC=-484.255, Time=0.23 sec
 ARIMA(3,1,2)(0,0,0)[0] intercept   : AIC=-485.081, Time=0.46 sec
 ARIMA(2,1,3)(0,0,0)[0] intercept   : AIC=-484.907, Time=0.45 sec
 ARIMA(1,1,1)(0,0,0)[0] intercept   : AIC=-484.833, Time=0.15 sec
 ARIMA(1,1,3)(0,0,0)[0] intercept   : AIC=-483.465, Time=0.53 sec
 ARIMA(3,1,1)(0,0,0)[0] intercept   : AIC=-482.253, Time=0.22 sec
 ARIMA(3,1,3)(0,0,0)[0] intercept   : AIC=-482.868, Time=0.52 sec
 ARIMA(2,1,2)(0,0,0)[0]             : AIC=-489.016, Time=0.18 sec
 ARIMA(1,1,2)(0,0,0)[0]             : AIC=-486.685, Time=0.11 sec
 ARIMA(2,1,1)(0,0,0)[0]             : AIC=-486.149, Time=0.07 sec
 ARIMA(3,1,2)(0,0,0)[0]             : AIC=-492.870, Time=0.25 sec
 ARIMA(3,1,1)(0,0,0)[0]             : AIC=-484.149, Time=0.10 sec
 ARIMA(4,1,2)(0,0,0)[0]             : AIC=-486.896, Time=0.29 sec
 ARIMA(3,1,3)(0,0,0)[0]             : AIC=-485.017, Time=0.32 sec
 ARIMA(2,1,3)(0,0,0)[0]             : AIC=-487.008, Time=0.25 sec
 ARIMA(4,1,1)(0,0,0)[0]             : AIC=-489.065, Time=0.30 sec
 ARIMA(4,1,3)(0,0,0)[0]             : AIC=inf, Time=0.41 sec

Best model:  ARIMA(3,1,2)(0,0,0)[0]          
Total fit time: 5.724 seconds
corn
Performing stepwise search to minimize aic
 ARIMA(2,1,2)(0,0,0)[0] intercept   : AIC=inf, Time=0.38 sec
 ARIMA(0,1,0)(0,0,0)[0] intercept   : AIC=115.242, Time=0.06 sec
 ARIMA(1,1,0)(0,0,0)[0] intercept   : AIC=105.876, Time=0.05 sec
 ARIMA(0,1,1)(0,0,0)[0] intercept   : AIC=106.856, Time=0.06 sec
 ARIMA(0,1,0)(0,0,0)[0]             : AIC=113.311, Time=0.03 sec
 ARIMA(2,1,0)(0,0,0)[0] intercept   : AIC=107.632, Time=0.07 sec
 ARIMA(1,1,1)(0,0,0)[0] intercept   : AIC=107.510, Time=0.09 sec
 ARIMA(2,1,1)(0,0,0)[0] intercept   : AIC=109.495, Time=0.16 sec
 ARIMA(1,1,0)(0,0,0)[0]             : AIC=103.909, Time=0.03 sec
 ARIMA(2,1,0)(0,0,0)[0]             : AIC=105.662, Time=0.03 sec
 ARIMA(1,1,1)(0,0,0)[0]             : AIC=105.538, Time=0.05 sec
 ARIMA(0,1,1)(0,0,0)[0]             : AIC=104.896, Time=0.03 sec
 ARIMA(2,1,1)(0,0,0)[0]             : AIC=107.523, Time=0.07 sec

Best model:  ARIMA(1,1,0)(0,0,0)[0]          
Total fit time: 1.117 seconds
coffee
Performing stepwise search to minimize aic
 ARIMA(2,1,2)(0,0,0)[0] intercept   : AIC=-362.400, Time=0.20 sec
 ARIMA(0,1,0)(0,0,0)[0] intercept   : AIC=-355.807, Time=0.04 sec
 ARIMA(1,1,0)(0,0,0)[0] intercept   : AIC=-364.728, Time=0.08 sec
 ARIMA(0,1,1)(0,0,0)[0] intercept   : AIC=-362.161, Time=0.08 sec
 ARIMA(0,1,0)(0,0,0)[0]             : AIC=-357.646, Time=0.03 sec
 ARIMA(2,1,0)(0,0,0)[0] intercept   : AIC=-366.085, Time=0.13 sec
 ARIMA(3,1,0)(0,0,0)[0] intercept   : AIC=-364.101, Time=0.12 sec
 ARIMA(2,1,1)(0,0,0)[0] intercept   : AIC=-364.096, Time=0.21 sec
 ARIMA(1,1,1)(0,0,0)[0] intercept   : AIC=-365.381, Time=0.13 sec
 ARIMA(3,1,1)(0,0,0)[0] intercept   : AIC=-362.093, Time=0.19 sec
 ARIMA(2,1,0)(0,0,0)[0]             : AIC=-368.013, Time=0.04 sec
 ARIMA(1,1,0)(0,0,0)[0]             : AIC=-366.632, Time=0.03 sec
 ARIMA(3,1,0)(0,0,0)[0]             : AIC=-366.028, Time=0.05 sec
 ARIMA(2,1,1)(0,0,0)[0]             : AIC=-366.024, Time=0.17 sec
 ARIMA(1,1,1)(0,0,0)[0]             : AIC=-367.315, Time=0.08 sec
 ARIMA(3,1,1)(0,0,0)[0]             : AIC=-364.021, Time=0.07 sec

Best model:  ARIMA(2,1,0)(0,0,0)[0]          
Total fit time: 1.664 seconds
soybean
Performing stepwise search to minimize aic
 ARIMA(2,1,2)(0,0,0)[0] intercept   : AIC=360.589, Time=0.19 sec
 ARIMA(0,1,0)(0,0,0)[0] intercept   : AIC=381.198, Time=0.04 sec
 ARIMA(1,1,0)(0,0,0)[0] intercept   : AIC=358.917, Time=0.05 sec
 ARIMA(0,1,1)(0,0,0)[0] intercept   : AIC=362.000, Time=0.05 sec
 ARIMA(0,1,0)(0,0,0)[0]             : AIC=379.365, Time=0.02 sec
 ARIMA(2,1,0)(0,0,0)[0] intercept   : AIC=360.741, Time=0.06 sec
 ARIMA(1,1,1)(0,0,0)[0] intercept   : AIC=360.822, Time=0.07 sec
 ARIMA(2,1,1)(0,0,0)[0] intercept   : AIC=358.764, Time=0.27 sec
 ARIMA(3,1,1)(0,0,0)[0] intercept   : AIC=360.573, Time=0.11 sec
 ARIMA(1,1,2)(0,0,0)[0] intercept   : AIC=359.278, Time=0.09 sec
 ARIMA(3,1,0)(0,0,0)[0] intercept   : AIC=358.663, Time=0.07 sec
 ARIMA(4,1,0)(0,0,0)[0] intercept   : AIC=360.552, Time=0.09 sec
 ARIMA(4,1,1)(0,0,0)[0] intercept   : AIC=362.550, Time=0.20 sec
 ARIMA(3,1,0)(0,0,0)[0]             : AIC=356.764, Time=0.03 sec
 ARIMA(2,1,0)(0,0,0)[0]             : AIC=358.808, Time=0.03 sec
 ARIMA(4,1,0)(0,0,0)[0]             : AIC=358.647, Time=0.04 sec
 ARIMA(3,1,1)(0,0,0)[0]             : AIC=358.670, Time=0.05 sec
 ARIMA(2,1,1)(0,0,0)[0]             : AIC=357.042, Time=0.11 sec
 ARIMA(4,1,1)(0,0,0)[0]             : AIC=360.645, Time=0.11 sec

Best model:  ARIMA(3,1,0)(0,0,0)[0]          
Total fit time: 1.695 seconds
soybean_oil
Performing stepwise search to minimize aic
 ARIMA(2,1,2)(0,0,0)[0] intercept   : AIC=-936.482, Time=0.41 sec
 ARIMA(0,1,0)(0,0,0)[0] intercept   : AIC=-902.384, Time=0.07 sec
 ARIMA(1,1,0)(0,0,0)[0] intercept   : AIC=-936.456, Time=0.09 sec
 ARIMA(0,1,1)(0,0,0)[0] intercept   : AIC=-927.959, Time=0.09 sec
 ARIMA(0,1,0)(0,0,0)[0]             : AIC=-904.303, Time=0.05 sec
 ARIMA(1,1,2)(0,0,0)[0] intercept   : AIC=-938.137, Time=0.30 sec
 ARIMA(0,1,2)(0,0,0)[0] intercept   : AIC=-936.369, Time=0.11 sec
 ARIMA(1,1,1)(0,0,0)[0] intercept   : AIC=-934.820, Time=0.10 sec
 ARIMA(1,1,3)(0,0,0)[0] intercept   : AIC=-932.371, Time=0.26 sec
 ARIMA(0,1,3)(0,0,0)[0] intercept   : AIC=-936.021, Time=0.28 sec
 ARIMA(2,1,1)(0,0,0)[0] intercept   : AIC=-934.986, Time=0.36 sec
 ARIMA(2,1,3)(0,0,0)[0] intercept   : AIC=-933.005, Time=0.43 sec
 ARIMA(1,1,2)(0,0,0)[0]             : AIC=-940.097, Time=0.14 sec
 ARIMA(0,1,2)(0,0,0)[0]             : AIC=-938.334, Time=0.06 sec
 ARIMA(1,1,1)(0,0,0)[0]             : AIC=-936.791, Time=0.04 sec
 ARIMA(2,1,2)(0,0,0)[0]             : AIC=-938.389, Time=0.20 sec
 ARIMA(1,1,3)(0,0,0)[0]             : AIC=-934.337, Time=0.12 sec
 ARIMA(0,1,1)(0,0,0)[0]             : AIC=-929.907, Time=0.04 sec
 ARIMA(0,1,3)(0,0,0)[0]             : AIC=-937.979, Time=0.09 sec
 ARIMA(2,1,1)(0,0,0)[0]             : AIC=-936.958, Time=0.17 sec
 ARIMA(2,1,3)(0,0,0)[0]             : AIC=-936.894, Time=0.27 sec

Best model:  ARIMA(1,1,2)(0,0,0)[0]          
Total fit time: 3.700 seconds
aluminum
Performing stepwise search to minimize aic
 ARIMA(2,0,2)(0,0,0)[0] intercept   : AIC=2296.659, Time=0.28 sec
 ARIMA(0,0,0)(0,0,0)[0] intercept   : AIC=2826.148, Time=0.02 sec
 ARIMA(1,0,0)(0,0,0)[0] intercept   : AIC=2306.074, Time=0.10 sec
 ARIMA(0,0,1)(0,0,0)[0] intercept   : AIC=inf, Time=0.07 sec
 ARIMA(0,0,0)(0,0,0)[0]             : AIC=3423.360, Time=0.02 sec
 ARIMA(1,0,2)(0,0,0)[0] intercept   : AIC=2294.308, Time=0.13 sec
 ARIMA(0,0,2)(0,0,0)[0] intercept   : AIC=2488.850, Time=0.28 sec
 ARIMA(1,0,1)(0,0,0)[0] intercept   : AIC=2293.298, Time=0.09 sec
 ARIMA(2,0,1)(0,0,0)[0] intercept   : AIC=2289.779, Time=0.19 sec
 ARIMA(2,0,0)(0,0,0)[0] intercept   : AIC=2291.300, Time=0.06 sec
 ARIMA(3,0,1)(0,0,0)[0] intercept   : AIC=2292.907, Time=0.36 sec
 ARIMA(3,0,0)(0,0,0)[0] intercept   : AIC=2292.474, Time=0.07 sec
 ARIMA(3,0,2)(0,0,0)[0] intercept   : AIC=inf, Time=0.41 sec
 ARIMA(2,0,1)(0,0,0)[0]             : AIC=2297.744, Time=0.10 sec

Best model:  ARIMA(2,0,1)(0,0,0)[0] intercept
Total fit time: 2.189 seconds

Aluminum ARIMA model¶

In [46]:
A_model_aluminum = ARIMA(all_prices_M_train['aluminum'], order = (3,1,3))
predictor_aluminum = A_model_aluminum.fit()
predictor_aluminum.summary()
Out[46]:
SARIMAX Results
Dep. Variable: aluminum No. Observations: 189
Model: ARIMA(3, 1, 3) Log Likelihood -1131.482
Date: Wed, 19 Apr 2023 AIC 2276.964
Time: 23:19:06 BIC 2299.619
Sample: 10-31-2002 HQIC 2286.143
- 06-30-2018
Covariance Type: opg
coef std err z P>|z| [0.025 0.975]
ar.L1 1.0153 0.239 4.241 0.000 0.546 1.484
ar.L2 -0.6343 0.370 -1.713 0.087 -1.360 0.092
ar.L3 0.1015 0.258 0.393 0.695 -0.405 0.608
ma.L1 -0.7859 0.265 -2.965 0.003 -1.305 -0.266
ma.L2 0.4842 0.365 1.325 0.185 -0.232 1.200
ma.L3 0.1629 0.236 0.690 0.490 -0.300 0.625
sigma2 9873.7916 783.905 12.596 0.000 8337.366 1.14e+04
Ljung-Box (L1) (Q): 0.00 Jarque-Bera (JB): 41.72
Prob(Q): 0.98 Prob(JB): 0.00
Heteroskedasticity (H): 0.41 Skew: -0.68
Prob(H) (two-sided): 0.00 Kurtosis: 4.87


Warnings:
[1] Covariance matrix calculated using the outer product of gradients (complex-step).
In [47]:
Predicted_results_aluminum = predictor_aluminum.predict(start= len(all_prices_M_train), end = (len(all_prices_M_train)+len(all_prices_M_test) - 1), typ = 'levels')
In [48]:
plt.plot(all_prices_M_test['aluminum'],color = 'red',label = 'Actual Values')
plt.plot(Predicted_results_aluminum,color = 'blue',label = 'Predicted Values')
plt.xlabel('Month')
plt.ylabel('Value')
plt.legend()
Out[48]:
<matplotlib.legend.Legend at 0x7f8e7c161190>
In [49]:
#performace metrics Soy bean oil ARIMA
RMSE_ARIMA_aluminum = math.sqrt(mean_squared_error(all_prices_M_test['aluminum'], Predicted_results_aluminum))
Rsquare_ARIMA_aluminum = r2_score(all_prices_M_test['aluminum'], Predicted_results_aluminum)
In [50]:
RMSE_ARIMA_aluminum
Out[50]:
505.2406577643067

Sugar ARIMA model¶

In [51]:
A_model_sugar = ARIMA(all_prices_M_train['sugar'], order = (6,0,7))
predictor_sugar = A_model_sugar.fit()
predictor_sugar.summary()

/usr/local/lib/python3.9/dist-packages/statsmodels/base/model.py:604: ConvergenceWarning: Maximum Likelihood optimization failed to converge. Check mle_retvals
  warnings.warn("Maximum Likelihood optimization failed to "
Out[51]:
SARIMAX Results
Dep. Variable: sugar No. Observations: 189
Model: ARIMA(6, 0, 7) Log Likelihood 556.174
Date: Wed, 19 Apr 2023 AIC -1082.348
Time: 23:19:08 BIC -1033.722
Sample: 10-31-2002 HQIC -1062.648
- 06-30-2018
Covariance Type: opg
coef std err z P>|z| [0.025 0.975]
const 0.1446 0.033 4.382 0.000 0.080 0.209
ar.L1 1.5690 1.131 1.387 0.165 -0.647 3.785
ar.L2 -1.0400 1.414 -0.736 0.462 -3.811 1.731
ar.L3 0.4557 1.482 0.308 0.758 -2.449 3.360
ar.L4 -0.4078 1.238 -0.329 0.742 -2.835 2.019
ar.L5 0.8901 0.991 0.898 0.369 -1.052 2.832
ar.L6 -0.4951 0.853 -0.580 0.562 -2.167 1.177
ma.L1 -0.0760 1.146 -0.066 0.947 -2.322 2.170
ma.L2 0.1074 0.826 0.130 0.897 -1.512 1.727
ma.L3 0.2182 0.869 0.251 0.802 -1.485 1.921
ma.L4 0.2626 0.745 0.352 0.725 -1.198 1.723
ma.L5 -0.2674 0.903 -0.296 0.767 -2.038 1.503
ma.L6 -0.2795 0.454 -0.615 0.538 -1.170 0.611
ma.L7 -0.0326 0.117 -0.278 0.781 -0.262 0.197
sigma2 0.0002 1.51e-05 10.590 0.000 0.000 0.000
Ljung-Box (L1) (Q): 0.22 Jarque-Bera (JB): 77.18
Prob(Q): 0.64 Prob(JB): 0.00
Heteroskedasticity (H): 2.01 Skew: 0.16
Prob(H) (two-sided): 0.01 Kurtosis: 6.11


Warnings:
[1] Covariance matrix calculated using the outer product of gradients (complex-step).
In [52]:
Predicted_results_sugar = predictor_sugar.predict(start= len(all_prices_M_train), end = (len(all_prices_M_train)+len(all_prices_M_test) - 1), typ = 'levels')
In [53]:
plt.plot(all_prices_M_test['sugar'],color = 'red',label = 'Actual Values')
plt.plot(Predicted_results_sugar,color = 'blue',label = 'Predicted Values')
plt.xlabel('Month')
plt.ylabel('Sugar Price')
plt.legend()
Out[53]:
<matplotlib.legend.Legend at 0x7f8e7c87aa90>
In [54]:
#performace metrics Sugar ARIMA
RMSE_ARIMA_sugar = math.sqrt(mean_squared_error(all_prices_M_test['sugar'], Predicted_results_sugar))
Rsquare_ARIMA_sugar = r2_score(all_prices_M_test['sugar'], Predicted_results_sugar)
In [55]:
RMSE_ARIMA_sugar
Out[55]:
0.02885330600594616

Cotton ARIMA model¶

In [56]:
A_model_cotton = ARIMA(all_prices_M_train['cotton'], order = (9,0,4))
predictor_cotton = A_model_cotton.fit()
predictor_cotton.summary()

/usr/local/lib/python3.9/dist-packages/statsmodels/base/model.py:604: ConvergenceWarning: Maximum Likelihood optimization failed to converge. Check mle_retvals
  warnings.warn("Maximum Likelihood optimization failed to "
Out[56]:
SARIMAX Results
Dep. Variable: cotton No. Observations: 189
Model: ARIMA(9, 0, 4) Log Likelihood 259.493
Date: Wed, 19 Apr 2023 AIC -488.986
Time: 23:19:09 BIC -440.360
Sample: 10-31-2002 HQIC -469.287
- 06-30-2018
Covariance Type: opg
coef std err z P>|z| [0.025 0.975]
const 0.7123 0.100 7.138 0.000 0.517 0.908
ar.L1 0.0412 1.509 0.027 0.978 -2.917 2.999
ar.L2 -0.0510 1.083 -0.047 0.962 -2.174 2.072
ar.L3 0.5126 0.532 0.963 0.335 -0.530 1.556
ar.L4 0.3230 0.655 0.493 0.622 -0.961 1.607
ar.L5 0.0994 0.503 0.198 0.843 -0.887 1.085
ar.L6 -0.2244 0.375 -0.598 0.550 -0.960 0.511
ar.L7 0.0879 0.196 0.448 0.654 -0.297 0.472
ar.L8 -0.0549 0.329 -0.167 0.867 -0.699 0.589
ar.L9 -0.0337 0.287 -0.117 0.907 -0.597 0.530
ma.L1 1.1518 1.520 0.758 0.449 -1.828 4.131
ma.L2 1.4003 1.008 1.388 0.165 -0.576 3.377
ma.L3 0.9346 1.245 0.751 0.453 -1.505 3.374
ma.L4 0.4399 0.689 0.639 0.523 -0.910 1.790
sigma2 0.0036 0.000 16.583 0.000 0.003 0.004
Ljung-Box (L1) (Q): 0.01 Jarque-Bera (JB): 993.09
Prob(Q): 0.94 Prob(JB): 0.00
Heteroskedasticity (H): 0.93 Skew: -0.53
Prob(H) (two-sided): 0.78 Kurtosis: 14.18


Warnings:
[1] Covariance matrix calculated using the outer product of gradients (complex-step).
In [57]:
Predicted_results_cotton = predictor_cotton.predict(start= len(all_prices_M_train), end = (len(all_prices_M_train)+len(all_prices_M_test) - 1), typ = 'levels')
In [58]:
plt.plot(all_prices_M_test['cotton'],color = 'red',label = 'Actual Values')
plt.plot(Predicted_results_cotton,color = 'blue',label = 'Predicted Values')
plt.xlabel('Month')
plt.ylabel('Cotton Price')
plt.legend()
Out[58]:
<matplotlib.legend.Legend at 0x7f8e7c1dfa30>
In [59]:
#performace metrics cotton ARIMA
RMSE_ARIMA_cotton = math.sqrt(mean_squared_error(all_prices_M_test['cotton'], Predicted_results_cotton))
Rsquare_ARIMA_cotton = r2_score(all_prices_M_test['cotton'], Predicted_results_cotton)
In [60]:
RMSE_ARIMA_cotton
Out[60]:
0.257799108563674

Corn ARIMA model¶

In [61]:
A_model_corn = ARIMA(all_prices_M_train['corn'], order = (3,1,5))
predictor_corn = A_model_corn.fit()
predictor_corn.summary()

/usr/local/lib/python3.9/dist-packages/statsmodels/base/model.py:604: ConvergenceWarning: Maximum Likelihood optimization failed to converge. Check mle_retvals
  warnings.warn("Maximum Likelihood optimization failed to "
Out[61]:
SARIMAX Results
Dep. Variable: corn No. Observations: 189
Model: ARIMA(3, 1, 5) Log Likelihood -42.525
Date: Wed, 19 Apr 2023 AIC 103.049
Time: 23:19:10 BIC 132.177
Sample: 10-31-2002 HQIC 114.851
- 06-30-2018
Covariance Type: opg
coef std err z P>|z| [0.025 0.975]
ar.L1 0.9804 0.066 14.890 0.000 0.851 1.109
ar.L2 0.6399 0.100 6.397 0.000 0.444 0.836
ar.L3 -0.8630 0.062 -13.993 0.000 -0.984 -0.742
ma.L1 -0.7229 0.769 -0.941 0.347 -2.229 0.783
ma.L2 -0.8932 1.307 -0.683 0.494 -3.455 1.668
ma.L3 0.8136 0.619 1.315 0.189 -0.399 2.026
ma.L4 0.0437 0.093 0.469 0.639 -0.139 0.227
ma.L5 0.0589 0.082 0.723 0.470 -0.101 0.219
sigma2 0.0905 0.065 1.399 0.162 -0.036 0.217
Ljung-Box (L1) (Q): 0.00 Jarque-Bera (JB): 149.19
Prob(Q): 1.00 Prob(JB): 0.00
Heteroskedasticity (H): 1.94 Skew: -0.10
Prob(H) (two-sided): 0.01 Kurtosis: 7.36


Warnings:
[1] Covariance matrix calculated using the outer product of gradients (complex-step).
In [62]:
Predicted_results_corn = predictor_corn.predict(start= len(all_prices_M_train), end = (len(all_prices_M_train)+len(all_prices_M_test) - 1), typ = 'levels')
In [63]:
plt.plot(all_prices_M_test['corn'],color = 'red',label = 'Actual Values')
plt.plot(Predicted_results_corn,color = 'blue',label = 'Predicted Values')
plt.xlabel('Month')
plt.ylabel('Corn Price')
plt.legend()
Out[63]:
<matplotlib.legend.Legend at 0x7f8e77ed2400>
In [64]:
#performace metrics corn ARIMA
RMSE_ARIMA_corn = math.sqrt(mean_squared_error(all_prices_M_test['corn'], Predicted_results_corn))
Rsquare_ARIMA_corn = r2_score(all_prices_M_test['corn'], Predicted_results_corn)
In [65]:
RMSE_ARIMA_corn
Out[65]:
1.8382735266248653

Coffee ARIMA model¶

In [66]:
A_model_coffee = ARIMA(all_prices_M_train['coffee'], order = (4,1,5))
predictor_coffee = A_model_coffee.fit()
predictor_coffee.summary()

/usr/local/lib/python3.9/dist-packages/statsmodels/base/model.py:604: ConvergenceWarning: Maximum Likelihood optimization failed to converge. Check mle_retvals
  warnings.warn("Maximum Likelihood optimization failed to "
Out[66]:
SARIMAX Results
Dep. Variable: coffee No. Observations: 189
Model: ARIMA(4, 1, 5) Log Likelihood 193.151
Date: Wed, 19 Apr 2023 AIC -366.302
Time: 23:19:11 BIC -333.938
Sample: 10-31-2002 HQIC -353.189
- 06-30-2018
Covariance Type: opg
coef std err z P>|z| [0.025 0.975]
ar.L1 -0.4600 0.261 -1.762 0.078 -0.972 0.052
ar.L2 0.8358 0.201 4.152 0.000 0.441 1.230
ar.L3 0.6721 0.203 3.304 0.001 0.273 1.071
ar.L4 -0.1218 0.242 -0.503 0.615 -0.596 0.352
ma.L1 0.7021 0.273 2.568 0.010 0.166 1.238
ma.L2 -0.5941 0.302 -1.968 0.049 -1.186 -0.002
ma.L3 -0.7212 0.212 -3.401 0.001 -1.137 -0.306
ma.L4 -0.1456 0.306 -0.476 0.634 -0.745 0.454
ma.L5 -0.2323 0.106 -2.187 0.029 -0.440 -0.024
sigma2 0.0072 0.001 10.722 0.000 0.006 0.008
Ljung-Box (L1) (Q): 0.01 Jarque-Bera (JB): 17.38
Prob(Q): 0.94 Prob(JB): 0.00
Heteroskedasticity (H): 2.09 Skew: 0.13
Prob(H) (two-sided): 0.00 Kurtosis: 4.47


Warnings:
[1] Covariance matrix calculated using the outer product of gradients (complex-step).
In [67]:
Predicted_results_coffee = predictor_coffee.predict(start= len(all_prices_M_train), end = (len(all_prices_M_train)+len(all_prices_M_test) - 1), typ = 'levels')
In [68]:
plt.plot(all_prices_M_test['coffee'],color = 'red',label = 'Actual Values')
plt.plot(Predicted_results_coffee,color = 'blue',label = 'Predicted Values')
plt.xlabel('Month')
plt.ylabel('Coffee Price')
plt.legend()
Out[68]:
<matplotlib.legend.Legend at 0x7f8e7c1946a0>
In [69]:
#performace metrics Soy bean oil ARIMA
RMSE_ARIMA_coffee = math.sqrt(mean_squared_error(all_prices_M_test['coffee'], Predicted_results_coffee))
Rsquare_ARIMA_coffee = r2_score(all_prices_M_test['coffee'], Predicted_results_coffee)
In [70]:
RMSE_ARIMA_coffee
Out[70]:
0.5086124942858565

Soybean ARIMA model¶

In [71]:
A_model_soybean = ARIMA(all_prices_M_train['soybean'], order = (2,1,6))
predictor_soybean = A_model_soybean.fit()
predictor_soybean.summary()
Out[71]:
SARIMAX Results
Dep. Variable: soybean No. Observations: 189
Model: ARIMA(2, 1, 6) Log Likelihood -172.426
Date: Wed, 19 Apr 2023 AIC 362.852
Time: 23:19:12 BIC 391.980
Sample: 10-31-2002 HQIC 374.654
- 06-30-2018
Covariance Type: opg
coef std err z P>|z| [0.025 0.975]
ar.L1 -0.0055 0.299 -0.018 0.985 -0.591 0.580
ar.L2 0.6920 0.252 2.741 0.006 0.197 1.187
ma.L1 0.3624 0.304 1.190 0.234 -0.234 0.959
ma.L2 -0.5639 0.241 -2.343 0.019 -1.036 -0.092
ma.L3 -0.3552 0.112 -3.181 0.001 -0.574 -0.136
ma.L4 -0.1781 0.097 -1.840 0.066 -0.368 0.012
ma.L5 0.0264 0.093 0.285 0.776 -0.155 0.208
ma.L6 -0.0408 0.099 -0.412 0.680 -0.235 0.153
sigma2 0.3658 0.029 12.795 0.000 0.310 0.422
Ljung-Box (L1) (Q): 0.00 Jarque-Bera (JB): 32.72
Prob(Q): 0.99 Prob(JB): 0.00
Heteroskedasticity (H): 1.44 Skew: -0.34
Prob(H) (two-sided): 0.15 Kurtosis: 4.93


Warnings:
[1] Covariance matrix calculated using the outer product of gradients (complex-step).
In [72]:
Predicted_results_soybean = predictor_soybean.predict(start= len(all_prices_M_train), end = (len(all_prices_M_train)+len(all_prices_M_test) - 1), typ = 'levels')
In [73]:
plt.plot(all_prices_M_test['soybean'],color = 'red',label = 'Actual Values')
plt.plot(Predicted_results_soybean,color = 'blue',label = 'Predicted Values')
plt.xlabel('Month')
plt.ylabel('Soybean Price')
plt.legend()
Out[73]:
<matplotlib.legend.Legend at 0x7f8e7c6b2340>
In [74]:
#performace metrics Soy bean oil ARIMA
RMSE_ARIMA_soybean = math.sqrt(mean_squared_error(all_prices_M_test['soybean'], Predicted_results_soybean))
Rsquare_ARIMA_soybean = r2_score(all_prices_M_test['soybean'], Predicted_results_soybean)
In [75]:
RMSE_ARIMA_soybean
Out[75]:
3.367017508721487

Soybean oil ARIMA model¶

In [76]:
A_model_soybean_oil = ARIMA(all_prices_M_train['soybean_oil'], order = (3,1,4))
predictor_soybean_oil = A_model_soybean_oil.fit()
predictor_soybean_oil.summary()

/usr/local/lib/python3.9/dist-packages/statsmodels/base/model.py:604: ConvergenceWarning: Maximum Likelihood optimization failed to converge. Check mle_retvals
  warnings.warn("Maximum Likelihood optimization failed to "
Out[76]:
SARIMAX Results
Dep. Variable: soybean_oil No. Observations: 189
Model: ARIMA(3, 1, 4) Log Likelihood 476.759
Date: Wed, 19 Apr 2023 AIC -937.517
Time: 23:19:13 BIC -911.626
Sample: 10-31-2002 HQIC -927.027
- 06-30-2018
Covariance Type: opg
coef std err z P>|z| [0.025 0.975]
ar.L1 0.4635 0.309 1.499 0.134 -0.143 1.070
ar.L2 -0.1148 0.206 -0.558 0.577 -0.518 0.288
ar.L3 -0.0926 0.268 -0.346 0.730 -0.618 0.432
ma.L1 -0.0302 0.302 -0.100 0.920 -0.621 0.561
ma.L2 0.1705 0.265 0.642 0.521 -0.350 0.691
ma.L3 0.0222 0.226 0.098 0.922 -0.420 0.464
ma.L4 0.3162 0.123 2.576 0.010 0.076 0.557
sigma2 0.0004 3.25e-05 11.220 0.000 0.000 0.000
Ljung-Box (L1) (Q): 0.01 Jarque-Bera (JB): 58.55
Prob(Q): 0.93 Prob(JB): 0.00
Heteroskedasticity (H): 0.80 Skew: -0.54
Prob(H) (two-sided): 0.37 Kurtosis: 5.51


Warnings:
[1] Covariance matrix calculated using the outer product of gradients (complex-step).
In [77]:
Predicted_results_soybean_oil = predictor_soybean_oil.predict(start= len(all_prices_M_train), end = (len(all_prices_M_train)+len(all_prices_M_test) - 1), typ = 'levels')
In [78]:
plt.plot(all_prices_M_test['soybean_oil'],color = 'red',label = 'Actual Values')
plt.plot(Predicted_results_soybean_oil,color = 'blue',label = 'Predicted Values')
plt.xlabel('Month')
plt.ylabel('Soybean Oil Price')
plt.legend()
Out[78]:
<matplotlib.legend.Legend at 0x7f8e77dfc520>
In [79]:
#performace metrics Soy bean oil ARIMA
RMSE_ARIMA_soybean_oil = math.sqrt(mean_squared_error(all_prices_M_test['soybean_oil'], Predicted_results_soybean_oil))
Rsquare_ARIMA_soybean_oil = r2_score(all_prices_M_test['soybean_oil'], Predicted_results_soybean_oil)
In [80]:
RMSE_ARIMA_soybean_oil
Out[80]:
0.21916021303429323

ARIMA Results Comments ¶

You can see from the charts that there is quite a flat line of predictions in comparison to the actual values. The large difference is in part because of the large increases in price due to the pandemic, the other is because it is a prediction based on moving averages. It may not be the most effective predictions, but they will be dependable (for good and bad) and easily understood.

The values of RMSE will be used below to compare to the LSTM model in my final results section.

LSTM Models ¶

Pre-processing and Setup for LTSM¶

Split dataframes are better for this model, producing the least amount of errors while building these models.

In [81]:
#Scaling the data is essential for creating this NN model. This will be reverted further along.
sc = MinMaxScaler(feature_range= (0,1))
sc2 = MinMaxScaler(feature_range= (-1,1))
In [82]:
#Parameter grid for LSTM hyperparpmeter Grid Search
param_grid = {
    'units': [50, 60, 70, 200, 300],
    'activation': ['relu', 'tanh'],
    'optimizer': ['adam', 'rmsprop']
}
In [83]:
def mae(y_true, predictions):
    y_true, predictions = np.array(y_true), np.array(predictions)
    return np.mean(np.abs(y_true - predictions))

LSTM Aluminum ¶

Individual Pre-Processing - Aluminum¶

In [84]:
aluminum_LSTM = aluminum.resample('M').mean()
In [85]:
#Test and train split
x7 = len(aluminum_LSTM)
size7 = int(x7 * 0.8)
index7 = x7 - size7
aluminum_LSTM_train = aluminum_LSTM.iloc[:size7,].values
aluminum_LSTM_test = aluminum_LSTM.iloc[size7:,].values
In [86]:
#Scaling
aluminum_LSTM_train_scaled = sc.fit_transform(aluminum_LSTM_train)
aluminum_LSTM_test_scaled = sc.fit_transform(aluminum_LSTM_test)
In [87]:
#Lists for X train and Y train
x_train_al = []
y_train_al = []
#Window size, 1 year
WS = 11
#Creating arrays
for i in range(WS, len(aluminum_LSTM_train_scaled)):
  x_train_al.append(aluminum_LSTM_train_scaled[i-WS:i,0:1])
  y_train_al.append(aluminum_LSTM_train_scaled[i,0])

x_train_al, y_train_al =np.array(x_train_al), np.array(y_train_al)

Grid Search for Aluminum Hyperparameters

In [88]:
def create_model(units=50, activation='relu', optimizer='adam'):
    model = Sequential()
    model.add(LSTM(units=units, activation=activation, input_shape=(x_train_al.shape[1], x_train_al.shape[2])))
    model.add(Dense(1))
    model.compile(loss='mean_squared_error', optimizer=optimizer)
    return model
In [89]:
model_al_tune = KerasRegressor(build_fn=create_model, verbose=0)
In [90]:
grid_al = GridSearchCV(estimator=model_al_tune, param_grid=param_grid, scoring='neg_mean_squared_error', cv=3)
In [91]:
grid_result_al = grid_al.fit(x_train_al, y_train_al)

WARNING:tensorflow:5 out of the last 9 calls to <function Model.make_predict_function.<locals>.predict_function at 0x7f8e6c47a0d0> triggered tf.function retracing. Tracing is expensive and the excessive number of tracings could be due to (1) creating @tf.function repeatedly in a loop, (2) passing tensors with different shapes, (3) passing Python objects instead of tensors. For (1), please define your @tf.function outside of the loop. For (2), @tf.function has reduce_retracing=True option that can avoid unnecessary retracing. For (3), please refer to https://www.tensorflow.org/guide/function#controlling_retracing and https://www.tensorflow.org/api_docs/python/tf/function for  more details.
WARNING:tensorflow:6 out of the last 11 calls to <function Model.make_predict_function.<locals>.predict_function at 0x7f8e6c1ccb80> triggered tf.function retracing. Tracing is expensive and the excessive number of tracings could be due to (1) creating @tf.function repeatedly in a loop, (2) passing tensors with different shapes, (3) passing Python objects instead of tensors. For (1), please define your @tf.function outside of the loop. For (2), @tf.function has reduce_retracing=True option that can avoid unnecessary retracing. For (3), please refer to https://www.tensorflow.org/guide/function#controlling_retracing and https://www.tensorflow.org/api_docs/python/tf/function for  more details.
In [92]:
print("Best: %f using %s" % (grid_result_al.best_score_, grid_result_al.best_params_))

Best: -0.023848 using {'activation': 'tanh', 'optimizer': 'adam', 'units': 200}

Developing LSTM model - Aluminum¶

In [93]:
#First Layer
model_al = Sequential()
model_al.add(LSTM(units = 200, activation = 'tanh', return_sequences = True, input_shape = (x_train_al.shape[1],1)))

#Dropout to avoid over fitting
model_al.add(Dropout(0.2))

#Next layer no input - first hidden layer
model_al.add(LSTM(units = 200, return_sequences = True))
model_al.add(Dropout(0.2))

#Second hidden layer
model_al.add(LSTM(units = 200, return_sequences = True))
model_al.add(Dropout(0.2))


#Last layer- no return
model_al.add(LSTM(units = 200))
model_al.add(Dropout(0.2))

#Output
model_al.add(Dense(units = 1))

#Optimizer 
model_al.compile(optimizer = 'rmsprop', loss = 'mean_squared_error')

#Fitting model
model_al.fit(x_train_al,y_train_al, epochs = 50, batch_size = 6)

Epoch 1/50
31/31 [==============================] - 9s 52ms/step - loss: 0.0630
Epoch 2/50
31/31 [==============================] - 1s 46ms/step - loss: 0.0507
Epoch 3/50
31/31 [==============================] - 1s 46ms/step - loss: 0.0452
Epoch 4/50
31/31 [==============================] - 1s 45ms/step - loss: 0.0415
Epoch 5/50
31/31 [==============================] - 1s 45ms/step - loss: 0.0377
Epoch 6/50
31/31 [==============================] - 1s 46ms/step - loss: 0.0356
Epoch 7/50
31/31 [==============================] - 1s 47ms/step - loss: 0.0354
Epoch 8/50
31/31 [==============================] - 1s 46ms/step - loss: 0.0297
Epoch 9/50
31/31 [==============================] - 1s 47ms/step - loss: 0.0304
Epoch 10/50
31/31 [==============================] - 1s 47ms/step - loss: 0.0215
Epoch 11/50
31/31 [==============================] - 1s 47ms/step - loss: 0.0220
Epoch 12/50
31/31 [==============================] - 1s 46ms/step - loss: 0.0221
Epoch 13/50
31/31 [==============================] - 1s 48ms/step - loss: 0.0167
Epoch 14/50
31/31 [==============================] - 1s 48ms/step - loss: 0.0168
Epoch 15/50
31/31 [==============================] - 1s 46ms/step - loss: 0.0156
Epoch 16/50
31/31 [==============================] - 1s 45ms/step - loss: 0.0144
Epoch 17/50
31/31 [==============================] - 1s 46ms/step - loss: 0.0140
Epoch 18/50
31/31 [==============================] - 1s 45ms/step - loss: 0.0135
Epoch 19/50
31/31 [==============================] - 1s 46ms/step - loss: 0.0109
Epoch 20/50
31/31 [==============================] - 1s 46ms/step - loss: 0.0128
Epoch 21/50
31/31 [==============================] - 1s 46ms/step - loss: 0.0099
Epoch 22/50
31/31 [==============================] - 1s 46ms/step - loss: 0.0099
Epoch 23/50
31/31 [==============================] - 1s 47ms/step - loss: 0.0095
Epoch 24/50
31/31 [==============================] - 1s 47ms/step - loss: 0.0088
Epoch 25/50
31/31 [==============================] - 1s 45ms/step - loss: 0.0082
Epoch 26/50
31/31 [==============================] - 1s 45ms/step - loss: 0.0087
Epoch 27/50
31/31 [==============================] - 1s 46ms/step - loss: 0.0084
Epoch 28/50
31/31 [==============================] - 1s 47ms/step - loss: 0.0083
Epoch 29/50
31/31 [==============================] - 1s 47ms/step - loss: 0.0078
Epoch 30/50
31/31 [==============================] - 1s 47ms/step - loss: 0.0072
Epoch 31/50
31/31 [==============================] - 2s 49ms/step - loss: 0.0072
Epoch 32/50
31/31 [==============================] - 1s 45ms/step - loss: 0.0085
Epoch 33/50
31/31 [==============================] - 1s 47ms/step - loss: 0.0067
Epoch 34/50
31/31 [==============================] - 2s 49ms/step - loss: 0.0069
Epoch 35/50
31/31 [==============================] - 1s 47ms/step - loss: 0.0063
Epoch 36/50
31/31 [==============================] - 1s 45ms/step - loss: 0.0073
Epoch 37/50
31/31 [==============================] - 1s 46ms/step - loss: 0.0076
Epoch 38/50
31/31 [==============================] - 1s 47ms/step - loss: 0.0061
Epoch 39/50
31/31 [==============================] - 1s 47ms/step - loss: 0.0081
Epoch 40/50
31/31 [==============================] - 1s 46ms/step - loss: 0.0077
Epoch 41/50
31/31 [==============================] - 1s 47ms/step - loss: 0.0070
Epoch 42/50
31/31 [==============================] - 1s 46ms/step - loss: 0.0067
Epoch 43/50
31/31 [==============================] - 1s 46ms/step - loss: 0.0063
Epoch 44/50
31/31 [==============================] - 1s 46ms/step - loss: 0.0065
Epoch 45/50
31/31 [==============================] - 1s 46ms/step - loss: 0.0071
Epoch 46/50
31/31 [==============================] - 1s 45ms/step - loss: 0.0066
Epoch 47/50
31/31 [==============================] - 1s 45ms/step - loss: 0.0069
Epoch 48/50
31/31 [==============================] - 1s 47ms/step - loss: 0.0068
Epoch 49/50
31/31 [==============================] - 1s 46ms/step - loss: 0.0067
Epoch 50/50
31/31 [==============================] - 1s 46ms/step - loss: 0.0068
Out[93]:
<keras.callbacks.History at 0x7f8e7793b190>
In [94]:
#Evlauating the number of Epochs
plt.plot(range(len(model_al.history.history['loss'])),model_al.history.history['loss'])
plt.xlabel('Epoch Number')
plt.ylabel('Loss')
plt.show()

Predictions - Aluminum¶

In [95]:
#Prediction setup
predict_test_al = [] 
batch_one_al = aluminum_LSTM_train_scaled[-WS:]
batch_new_al = batch_one_al.reshape((1,WS,1))

#Set number of predictions in range. I have set 49 months which is the length of the test data
for i in range(49):

  first_pred = model_al.predict(batch_new_al)[0]

  predict_test_al.append(first_pred)

  batch_new_al = np.append(batch_new_al[:,1:,:], [[first_pred]], axis = 1)

predict_test_al = np.array(predict_test_al)

predictions_al = sc.inverse_transform(predict_test_al[:49])

1/1 [==============================] - 1s 1s/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step

Results - Aluminum¶

In [96]:
#Plot comparison between predicted and actual values
plt.plot(aluminum_LSTM_test,color = 'red', label = 'Actual Values')
plt.plot(predictions_al, color = 'blue', label = 'Predicted Values')
plt.title('Predicted VS Actual Values')
plt.xlabel('Time (M)')
plt.ylabel('Aluminum Price')
plt.legend()
plt.show()
In [97]:
#Performace metrics
Aluminum_RMSE = math.sqrt(mean_squared_error(aluminum_LSTM_test, predictions_al))
In [98]:
Aluminum_RMSE
Out[98]:
502.53211929570585
In [99]:
mae(aluminum_LSTM_test, predictions_al)
Out[99]:
425.33802130979285

LSTM Sugar ¶

Individual Pre-Processing for Sugar¶

In [100]:
sugar_LSTM = sugar.resample('M').mean()
In [101]:
x = len(sugar_LSTM)
size = int(x * 0.8)
index = x - size
sugar_LSTM_train = sugar_LSTM.iloc[:size,].values
sugar_LSTM_test = sugar_LSTM.iloc[size:,].values
In [102]:
sugar_LSTM_train_scaled = sc2.fit_transform(sugar_LSTM_train)
sugar_LSTM_test_scaled = sc2.fit_transform(sugar_LSTM_test)
In [103]:
#Lists for X train and Y train
x_train_su = []
y_train_su = []
#Window size
WS = 12
#creating arrays for model
for i in range(WS, len(sugar_LSTM_train_scaled)):
  x_train_su.append(sugar_LSTM_train_scaled[i-WS:i,0:1])
  y_train_su.append(sugar_LSTM_train_scaled[i,0])

x_train_su, y_train_su =np.array(x_train_su), np.array(y_train_su)

Grid Search for Sugar Hyperparameters

In [104]:
def create_model(units=50, activation='relu', optimizer='adam'):
    model = Sequential()
    model.add(LSTM(units=units, activation=activation, input_shape=(x_train_su.shape[1], x_train_su.shape[2])))
    model.add(Dense(1))
    model.compile(loss='mean_squared_error', optimizer=optimizer)
    return model
In [105]:
model_su_tune = KerasRegressor(build_fn=create_model, verbose=0)
In [106]:
grid_su = GridSearchCV(estimator=model_su_tune, param_grid=param_grid, scoring='neg_mean_squared_error', cv=3)
In [107]:
grid_result_su = grid_su.fit(x_train_su, y_train_su)
In [108]:
print("Best: %f using %s" % (grid_result_su.best_score_, grid_result_su.best_params_))

Best: -0.022491 using {'activation': 'relu', 'optimizer': 'rmsprop', 'units': 200}

Developing LSTM model - Sugar¶

In [109]:
#First Layer
model_su = Sequential()
model_su.add(LSTM(units = 200, activation = 'relu', return_sequences = True, input_shape = (x_train_su.shape[1],1)))

#Dropout to avoid over fitting
model_su.add(Dropout(0.2))

#Next layer no input - first hidden layer
model_su.add(LSTM(units = 200, return_sequences = True))
model_su.add(Dropout(0.2))

#Second hidden layer
model_su.add(LSTM(units = 200, return_sequences = True))
model_su.add(Dropout(0.2))

#Last layer- no return
model_su.add(LSTM(units = 200))
model_su.add(Dropout(0.2))

#Output
model_su.add(Dense(units = 1))

#Optimizer 
model_su.compile(optimizer = 'rmsprop', loss = 'mean_squared_error')

#Fitting model
model_su.fit(x_train_su,y_train_su, epochs = 50, batch_size = 5)

Epoch 1/50
114/114 [==============================] - 12s 49ms/step - loss: 0.0751
Epoch 2/50
114/114 [==============================] - 6s 49ms/step - loss: 0.0462
Epoch 3/50
114/114 [==============================] - 6s 49ms/step - loss: 0.0376
Epoch 4/50
114/114 [==============================] - 6s 49ms/step - loss: 0.0303
Epoch 5/50
114/114 [==============================] - 6s 49ms/step - loss: 0.0244
Epoch 6/50
114/114 [==============================] - 5s 48ms/step - loss: 0.0206
Epoch 7/50
114/114 [==============================] - 6s 49ms/step - loss: 0.0171
Epoch 8/50
114/114 [==============================] - 6s 50ms/step - loss: 0.0169
Epoch 9/50
114/114 [==============================] - 6s 49ms/step - loss: 0.0147
Epoch 10/50
114/114 [==============================] - 6s 50ms/step - loss: 0.0132
Epoch 11/50
114/114 [==============================] - 6s 49ms/step - loss: 0.0117
Epoch 12/50
114/114 [==============================] - 6s 50ms/step - loss: 0.0109
Epoch 13/50
114/114 [==============================] - 6s 49ms/step - loss: 0.0108
Epoch 14/50
114/114 [==============================] - 6s 49ms/step - loss: 0.0098
Epoch 15/50
114/114 [==============================] - 6s 50ms/step - loss: 0.0097
Epoch 16/50
114/114 [==============================] - 6s 49ms/step - loss: 0.0083
Epoch 17/50
114/114 [==============================] - 6s 51ms/step - loss: 0.0086
Epoch 18/50
114/114 [==============================] - 6s 50ms/step - loss: 0.0085
Epoch 19/50
114/114 [==============================] - 6s 49ms/step - loss: 0.0081
Epoch 20/50
114/114 [==============================] - 5s 48ms/step - loss: 0.0078
Epoch 21/50
114/114 [==============================] - 6s 49ms/step - loss: 0.0100
Epoch 22/50
114/114 [==============================] - 6s 49ms/step - loss: 0.0080
Epoch 23/50
114/114 [==============================] - 6s 49ms/step - loss: 0.0079
Epoch 24/50
114/114 [==============================] - 6s 52ms/step - loss: 0.0074
Epoch 25/50
114/114 [==============================] - 6s 50ms/step - loss: 0.0081
Epoch 26/50
114/114 [==============================] - 6s 48ms/step - loss: 0.0073
Epoch 27/50
114/114 [==============================] - 6s 50ms/step - loss: 0.0078
Epoch 28/50
114/114 [==============================] - 6s 48ms/step - loss: 0.0073
Epoch 29/50
114/114 [==============================] - 5s 48ms/step - loss: 0.0071
Epoch 30/50
114/114 [==============================] - 6s 49ms/step - loss: 0.0068
Epoch 31/50
114/114 [==============================] - 6s 49ms/step - loss: 0.0081
Epoch 32/50
114/114 [==============================] - 6s 51ms/step - loss: 0.0069
Epoch 33/50
114/114 [==============================] - 6s 48ms/step - loss: 0.0070
Epoch 34/50
114/114 [==============================] - 6s 49ms/step - loss: 0.0068
Epoch 35/50
114/114 [==============================] - 6s 50ms/step - loss: 0.0071
Epoch 36/50
114/114 [==============================] - 5s 48ms/step - loss: 0.0074
Epoch 37/50
114/114 [==============================] - 6s 50ms/step - loss: 0.0067
Epoch 38/50
114/114 [==============================] - 5s 48ms/step - loss: 0.0068
Epoch 39/50
114/114 [==============================] - 6s 49ms/step - loss: 0.0061
Epoch 40/50
114/114 [==============================] - 5s 48ms/step - loss: 0.0069
Epoch 41/50
114/114 [==============================] - 5s 47ms/step - loss: 0.0068
Epoch 42/50
114/114 [==============================] - 5s 48ms/step - loss: 0.0061
Epoch 43/50
114/114 [==============================] - 6s 48ms/step - loss: 0.0055
Epoch 44/50
114/114 [==============================] - 6s 50ms/step - loss: 0.0073
Epoch 45/50
114/114 [==============================] - 6s 49ms/step - loss: 0.0063
Epoch 46/50
114/114 [==============================] - 5s 48ms/step - loss: 0.0051
Epoch 47/50
114/114 [==============================] - 6s 49ms/step - loss: 0.0058
Epoch 48/50
114/114 [==============================] - 6s 50ms/step - loss: 0.0062
Epoch 49/50
114/114 [==============================] - 5s 48ms/step - loss: 0.0058
Epoch 50/50
114/114 [==============================] - 6s 54ms/step - loss: 0.0065
Out[109]:
<keras.callbacks.History at 0x7f8e644f9dc0>
In [110]:
#Plotting loss in Epochs, determining if we are using the correct amount of epochs
plt.plot(range(len(model_su.history.history['loss'])),model_su.history.history['loss'])
plt.xlabel('Epoch Number')
plt.ylabel('Loss')
plt.show()

Predictions - Sugar¶

In [111]:
#Prediction setup
predict_test_su = []
batch_one_su = sugar_LSTM_train_scaled[-WS:]
batch_new_su = batch_one_su.reshape((1,WS,1))

#Set number of predictions in range.
for i in range(len(sugar_LSTM_test_scaled)):

  first_pred = model_su.predict(batch_new_su)[0]

  predict_test_su.append(first_pred)

  batch_new_su = np.append(batch_new_su[:,1:,:], [[first_pred]], axis = 1)

1/1 [==============================] - 2s 2s/step
1/1 [==============================] - 0s 32ms/step
1/1 [==============================] - 0s 32ms/step
1/1 [==============================] - 0s 32ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 33ms/step
1/1 [==============================] - 0s 32ms/step
1/1 [==============================] - 0s 32ms/step
1/1 [==============================] - 0s 33ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 32ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 32ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 32ms/step
1/1 [==============================] - 0s 32ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 32ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 35ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 33ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 39ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 33ms/step
1/1 [==============================] - 0s 34ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 33ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 32ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 32ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 29ms/step
In [112]:
#changing to array
predict_test_su = np.array(predict_test_su)
#unscaling the values
predictions_su = sc2.inverse_transform(predict_test_su)

Results - Sugar¶

In [113]:
#Plot comparison between predicted and actual values
plt.plot(sugar_LSTM_test,color = 'red', label = 'Actual Values')
plt.plot(predictions_su, color = 'blue', label = 'Predicted Values')
plt.title('Predicted VS Actual Values')
plt.xlabel('Time (M)')
plt.ylabel('Sugar Price')
plt.legend()
plt.show()
In [114]:
#Performace metrics
RMSE_sugar = math.sqrt(mean_squared_error(sugar_LSTM_test, predictions_su))
Rsquare_sugar = r2_score(sugar_LSTM_test, predictions_su)
In [115]:
RMSE_sugar
Out[115]:
0.042438750327986836
In [116]:
mae(sugar_LSTM_test, predictions_su)
Out[116]:
0.033755289349415026

LSTM Cotton ¶

Individual Pre-Processing - Cotton¶

In [117]:
cotton_LSTM = cotton.resample('M').mean()
In [118]:
#Test train Split
x2 = len(cotton_LSTM)
size2 = int(x2 * 0.8)
index2 = x2 - size2
cotton_LSTM_train = cotton_LSTM.iloc[:size2,].values
cotton_LSTM_test = cotton_LSTM.iloc[size2:,].values
In [119]:
cotton_LSTM_train_scaled = sc2.fit_transform(cotton_LSTM_train)
cotton_LSTM_test_scaled = sc2.fit_transform(cotton_LSTM_test)
In [120]:
#Lists for X and Y train
x_train_cot = []
y_train_cot = []
#Window size
WS = 10
#creating arrays for model
for i in range(WS, len(cotton_LSTM_train_scaled)):
  x_train_cot.append(cotton_LSTM_train_scaled[i-WS:i,0:1])
  y_train_cot.append(cotton_LSTM_train_scaled[i,0])

x_train_cot, y_train_cot =np.array(x_train_cot), np.array(y_train_cot)

Gridsearch for Cotton Hyperparameters

In [121]:
def create_model(units=50, activation='relu', optimizer='adam'):
    model = Sequential()
    model.add(LSTM(units=units, activation=activation, input_shape=(x_train_cot.shape[1], x_train_cot.shape[2])))
    model.add(Dense(1))
    model.compile(loss='mean_squared_error', optimizer=optimizer)
    return model
In [122]:
model_cot_tune = KerasRegressor(build_fn=create_model, verbose=0)
In [123]:
grid_cot = GridSearchCV(estimator=model_cot_tune, param_grid=param_grid, scoring='neg_mean_squared_error', cv=3)
In [124]:
grid_result_cot = grid_cot.fit(x_train_cot, y_train_cot)
In [125]:
print("Best: %f using %s" % (grid_result_cot.best_score_, grid_result_cot.best_params_))

Best: -0.021247 using {'activation': 'tanh', 'optimizer': 'rmsprop', 'units': 50}

Developing LSTM Model - Cotton¶

In [126]:
#First Layer
model_cot = Sequential()
model_cot.add(LSTM(units = 50, activation = 'tanh', return_sequences = True, input_shape = (x_train_cot.shape[1],1)))

#Dropout to avoid over fitting
model_cot.add(Dropout(0.2))

#Next layer no input - first hidden layer
model_cot.add(LSTM(units = 50, return_sequences = True))
model_cot.add(Dropout(0.2))

#Second hidden layer
model_cot.add(LSTM(units = 50, return_sequences = True))
model_cot.add(Dropout(0.2))

#Last layer- no return
model_cot.add(LSTM(units = 50))
model_cot.add(Dropout(0.2))

#Output
model_cot.add(Dense(units = 1))

#Optimizer
model_cot.compile(optimizer = 'rmsprop', loss = 'mean_squared_error')

#fitting model
model_cot.fit(x_train_cot,y_train_cot, epochs = 50, batch_size = 6)

Epoch 1/50
80/80 [==============================] - 10s 19ms/step - loss: 0.0549
Epoch 2/50
80/80 [==============================] - 1s 18ms/step - loss: 0.0404
Epoch 3/50
80/80 [==============================] - 1s 18ms/step - loss: 0.0377
Epoch 4/50
80/80 [==============================] - 1s 18ms/step - loss: 0.0373
Epoch 5/50
80/80 [==============================] - 1s 19ms/step - loss: 0.0351
Epoch 6/50
80/80 [==============================] - 2s 19ms/step - loss: 0.0321
Epoch 7/50
80/80 [==============================] - 1s 19ms/step - loss: 0.0292
Epoch 8/50
80/80 [==============================] - 1s 19ms/step - loss: 0.0270
Epoch 9/50
80/80 [==============================] - 1s 18ms/step - loss: 0.0251
Epoch 10/50
80/80 [==============================] - 1s 18ms/step - loss: 0.0223
Epoch 11/50
80/80 [==============================] - 1s 19ms/step - loss: 0.0215
Epoch 12/50
80/80 [==============================] - 1s 18ms/step - loss: 0.0190
Epoch 13/50
80/80 [==============================] - 1s 18ms/step - loss: 0.0180
Epoch 14/50
80/80 [==============================] - 2s 19ms/step - loss: 0.0166
Epoch 15/50
80/80 [==============================] - 2s 19ms/step - loss: 0.0156
Epoch 16/50
80/80 [==============================] - 2s 19ms/step - loss: 0.0149
Epoch 17/50
80/80 [==============================] - 1s 18ms/step - loss: 0.0142
Epoch 18/50
80/80 [==============================] - 1s 18ms/step - loss: 0.0127
Epoch 19/50
80/80 [==============================] - 1s 19ms/step - loss: 0.0119
Epoch 20/50
80/80 [==============================] - 2s 19ms/step - loss: 0.0122
Epoch 21/50
80/80 [==============================] - 1s 18ms/step - loss: 0.0121
Epoch 22/50
80/80 [==============================] - 1s 19ms/step - loss: 0.0099
Epoch 23/50
80/80 [==============================] - 1s 18ms/step - loss: 0.0090
Epoch 24/50
80/80 [==============================] - 1s 18ms/step - loss: 0.0086
Epoch 25/50
80/80 [==============================] - 2s 19ms/step - loss: 0.0086
Epoch 26/50
80/80 [==============================] - 2s 19ms/step - loss: 0.0085
Epoch 27/50
80/80 [==============================] - 1s 18ms/step - loss: 0.0090
Epoch 28/50
80/80 [==============================] - 1s 18ms/step - loss: 0.0081
Epoch 29/50
80/80 [==============================] - 1s 18ms/step - loss: 0.0080
Epoch 30/50
80/80 [==============================] - 1s 18ms/step - loss: 0.0088
Epoch 31/50
80/80 [==============================] - 1s 18ms/step - loss: 0.0083
Epoch 32/50
80/80 [==============================] - 1s 18ms/step - loss: 0.0081
Epoch 33/50
80/80 [==============================] - 1s 18ms/step - loss: 0.0084
Epoch 34/50
80/80 [==============================] - 1s 18ms/step - loss: 0.0079
Epoch 35/50
80/80 [==============================] - 2s 19ms/step - loss: 0.0070
Epoch 36/50
80/80 [==============================] - 1s 19ms/step - loss: 0.0071
Epoch 37/50
80/80 [==============================] - 1s 18ms/step - loss: 0.0071
Epoch 38/50
80/80 [==============================] - 2s 19ms/step - loss: 0.0073
Epoch 39/50
80/80 [==============================] - 2s 19ms/step - loss: 0.0069
Epoch 40/50
80/80 [==============================] - 1s 19ms/step - loss: 0.0075
Epoch 41/50
80/80 [==============================] - 1s 19ms/step - loss: 0.0070
Epoch 42/50
80/80 [==============================] - 2s 19ms/step - loss: 0.0063
Epoch 43/50
80/80 [==============================] - 1s 18ms/step - loss: 0.0068
Epoch 44/50
80/80 [==============================] - 1s 18ms/step - loss: 0.0077
Epoch 45/50
80/80 [==============================] - 1s 19ms/step - loss: 0.0069
Epoch 46/50
80/80 [==============================] - 1s 19ms/step - loss: 0.0056
Epoch 47/50
80/80 [==============================] - 1s 18ms/step - loss: 0.0065
Epoch 48/50
80/80 [==============================] - 1s 18ms/step - loss: 0.0062
Epoch 49/50
80/80 [==============================] - 1s 18ms/step - loss: 0.0060
Epoch 50/50
80/80 [==============================] - 2s 19ms/step - loss: 0.0068
Out[126]:
<keras.callbacks.History at 0x7f8d80616370>
In [127]:
#Evlauating the number of Epochs
plt.plot(range(len(model_cot.history.history['loss'])),model_cot.history.history['loss'])
plt.xlabel('Epoch Number')
plt.ylabel('Loss')
plt.show()

Predictions - Cotton¶

In [128]:
#Prediction setup
predict_test_cot = [] 
batch_one_cot = cotton_LSTM_train_scaled[-WS:]
batch_new_cot = batch_one_cot.reshape((1,WS,1))

#Set number of predictions in range.
for i in range(len(cotton_LSTM_test_scaled)):

  first_pred = model_cot.predict(batch_new_cot)[0]

  predict_test_cot.append(first_pred)

  batch_new_cot = np.append(batch_new_cot[:,1:,:], [[first_pred]], axis = 1)

1/1 [==============================] - 2s 2s/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 26ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 26ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 26ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 26ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 26ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 26ms/step
1/1 [==============================] - 0s 26ms/step
1/1 [==============================] - 0s 26ms/step
1/1 [==============================] - 0s 26ms/step
1/1 [==============================] - 0s 26ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 26ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 26ms/step
1/1 [==============================] - 0s 25ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 26ms/step
1/1 [==============================] - 0s 26ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 26ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 26ms/step
1/1 [==============================] - 0s 32ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 26ms/step
1/1 [==============================] - 0s 26ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 26ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 26ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 26ms/step
1/1 [==============================] - 0s 26ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 26ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 25ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
In [129]:
#Changing to array
predict_test_cot = np.array(predict_test_cot)
#Unscaling values
predictions_cot = sc2.inverse_transform(predict_test_cot)

Results for Cotton¶

In [130]:
#Plot comparison between predicted and actual values
plt.plot(cotton_LSTM_test,color = 'red', label = 'Actual Values')
plt.plot(predictions_cot, color = 'blue', label = 'Predicted Values')
plt.title('Predicted VS Actual Values')
plt.xlabel('Time (M)')
plt.ylabel('Cotton Price')
plt.legend()
plt.show()
In [131]:
#Performace metrics
RMSE_cotton = math.sqrt(mean_squared_error(cotton_LSTM_test, predictions_cot))
In [132]:
RMSE_cotton
Out[132]:
0.7173651109058645
In [133]:
mae(cotton_LSTM_test, predictions_cot)
Out[133]:
0.6718357316967694

LSTM Corn ¶

Individual pre-processing for Corn¶

In [134]:
corn_LSTM = corn.resample('M').mean()
In [135]:
x3 = len(corn_LSTM)
size3 = int(x3 * 0.8)
index3 = x3 - size3
corn_LSTM_train = corn_LSTM.iloc[:size3,].values
corn_LSTM_test = corn_LSTM.iloc[size3:,].values
In [136]:
corn_LSTM_train_scaled = sc2.fit_transform(corn_LSTM_train)
corn_LSTM_test_scaled = sc2.fit_transform(corn_LSTM_test)
In [137]:
#Lists for X train and Y train
x_train_cor = []
y_train_cor = []
#Window Size
WS = 12
#Creating arrays for model
for i in range(WS,len(corn_LSTM_train_scaled)):
  x_train_cor.append(corn_LSTM_train_scaled[i-WS:i,0:1])
  y_train_cor.append(corn_LSTM_train_scaled[i,0])

x_train_cor, y_train_cor =np.array(x_train_cor), np.array(y_train_cor)

Gridsearch for Corn Hyperparameters

In [138]:
def create_model(units=50, activation='relu', optimizer='adam'):
    model = Sequential()
    model.add(LSTM(units=units, activation=activation, input_shape=(x_train_cor.shape[1], x_train_cor.shape[2])))
    model.add(Dense(1))
    model.compile(loss='mean_squared_error', optimizer=optimizer)
    return model
In [139]:
model_cor_tune = KerasRegressor(build_fn=create_model, verbose=0)
In [140]:
grid_cor = GridSearchCV(estimator=model_cor_tune, param_grid=param_grid, scoring='neg_mean_squared_error', cv=3)
In [141]:
grid_result_cor = grid_cor.fit(x_train_cor, y_train_cor)
In [142]:
print("Best: %f using %s" % (grid_result_cor.best_score_, grid_result_cor.best_params_))

Best: -0.022904 using {'activation': 'tanh', 'optimizer': 'rmsprop', 'units': 300}

Developing LSTM model for Corn¶

In [143]:
#First layer
model_cor = Sequential()
model_cor.add(LSTM(units = 200, activation = 'relu', return_sequences = True, input_shape = (x_train_cor.shape[1],1)))

#Dropout to avoid over fitting
model_cor.add(Dropout(0.2))

#Next layer no input - first hidden layer
model_cor.add(LSTM(units = 200, return_sequences = True))
model_cor.add(Dropout(0.2))

#Second hidden layer
model_cor.add(LSTM(units = 200, return_sequences = True))
model_cor.add(Dropout(0.2))

#Last layer- no return
model_cor.add(LSTM(units = 200))
model_cor.add(Dropout(0.2))

#Output
model_cor.add(Dense(units = 1))

#Optimizer
model_cor.compile(optimizer = 'rmsprop', loss = 'mean_squared_error')

#Fitting model
model_cor.fit(x_train_cor,y_train_cor, epochs = 50, batch_size = 6)

Epoch 1/50
100/100 [==============================] - 12s 51ms/step - loss: 0.0695
Epoch 2/50
100/100 [==============================] - 5s 51ms/step - loss: 0.0406
Epoch 3/50
100/100 [==============================] - 5s 51ms/step - loss: 0.0334
Epoch 4/50
100/100 [==============================] - 5s 52ms/step - loss: 0.0288
Epoch 5/50
100/100 [==============================] - 5s 50ms/step - loss: 0.0271
Epoch 6/50
100/100 [==============================] - 5s 50ms/step - loss: 0.0216
Epoch 7/50
100/100 [==============================] - 5s 50ms/step - loss: 0.0181
Epoch 8/50
100/100 [==============================] - 5s 49ms/step - loss: 0.0161
Epoch 9/50
100/100 [==============================] - 5s 51ms/step - loss: 0.0161
Epoch 10/50
100/100 [==============================] - 5s 50ms/step - loss: 0.0145
Epoch 11/50
100/100 [==============================] - 5s 50ms/step - loss: 0.0120
Epoch 12/50
100/100 [==============================] - 5s 53ms/step - loss: 0.0112
Epoch 13/50
100/100 [==============================] - 5s 50ms/step - loss: 0.0101
Epoch 14/50
100/100 [==============================] - 5s 49ms/step - loss: 0.0105
Epoch 15/50
100/100 [==============================] - 5s 51ms/step - loss: 0.0097
Epoch 16/50
100/100 [==============================] - 5s 49ms/step - loss: 0.0098
Epoch 17/50
100/100 [==============================] - 5s 50ms/step - loss: 0.0091
Epoch 18/50
100/100 [==============================] - 5s 51ms/step - loss: 0.0091
Epoch 19/50
100/100 [==============================] - 5s 50ms/step - loss: 0.0086
Epoch 20/50
100/100 [==============================] - 5s 51ms/step - loss: 0.0091
Epoch 21/50
100/100 [==============================] - 5s 51ms/step - loss: 0.0079
Epoch 22/50
100/100 [==============================] - 5s 49ms/step - loss: 0.0080
Epoch 23/50
100/100 [==============================] - 5s 50ms/step - loss: 0.0078
Epoch 24/50
100/100 [==============================] - 5s 50ms/step - loss: 0.0071
Epoch 25/50
100/100 [==============================] - 5s 49ms/step - loss: 0.0070
Epoch 26/50
100/100 [==============================] - 5s 52ms/step - loss: 0.0070
Epoch 27/50
100/100 [==============================] - 5s 52ms/step - loss: 0.0074
Epoch 28/50
100/100 [==============================] - 5s 52ms/step - loss: 0.0064
Epoch 29/50
100/100 [==============================] - 5s 52ms/step - loss: 0.0083
Epoch 30/50
100/100 [==============================] - 5s 52ms/step - loss: 0.0068
Epoch 31/50
100/100 [==============================] - 5s 52ms/step - loss: 0.0071
Epoch 32/50
100/100 [==============================] - 5s 52ms/step - loss: 0.0071
Epoch 33/50
100/100 [==============================] - 5s 51ms/step - loss: 0.0076
Epoch 34/50
100/100 [==============================] - 5s 52ms/step - loss: 0.0067
Epoch 35/50
100/100 [==============================] - 5s 51ms/step - loss: 0.0066
Epoch 36/50
100/100 [==============================] - 5s 54ms/step - loss: 0.0077
Epoch 37/50
100/100 [==============================] - 6s 56ms/step - loss: 0.0062
Epoch 38/50
100/100 [==============================] - 5s 52ms/step - loss: 0.0074
Epoch 39/50
100/100 [==============================] - 5s 52ms/step - loss: 0.0065
Epoch 40/50
100/100 [==============================] - 5s 50ms/step - loss: 0.0061
Epoch 41/50
100/100 [==============================] - 5s 50ms/step - loss: 0.0059
Epoch 42/50
100/100 [==============================] - 5s 53ms/step - loss: 0.0060
Epoch 43/50
100/100 [==============================] - 5s 54ms/step - loss: 0.0057
Epoch 44/50
100/100 [==============================] - 5s 51ms/step - loss: 0.0071
Epoch 45/50
100/100 [==============================] - 5s 52ms/step - loss: 0.0066
Epoch 46/50
100/100 [==============================] - 5s 51ms/step - loss: 0.0072
Epoch 47/50
100/100 [==============================] - 5s 51ms/step - loss: 0.0069
Epoch 48/50
100/100 [==============================] - 5s 53ms/step - loss: 0.0063
Epoch 49/50
100/100 [==============================] - 5s 52ms/step - loss: 0.0067
Epoch 50/50
100/100 [==============================] - 5s 52ms/step - loss: 0.0065
Out[143]:
<keras.callbacks.History at 0x7f8dc4111a30>
In [144]:
#Evaluating the number of epochs
plt.plot(range(len(model_cor.history.history['loss'])),model_cor.history.history['loss'])
plt.xlabel('Epoch Number')
plt.ylabel('Loss')
plt.show()

Predictions for Corn¶

In [145]:
#Prediction setup
predict_test_cor = [] 
batch_one_cor = corn_LSTM_train_scaled[-WS:]
batch_new_cor = batch_one_cor.reshape((1,WS,1))

#Set number of predictions in range
for i in range(len(corn_LSTM_test_scaled)):

  first_pred = model_cor.predict(batch_new_cor)[0]

  predict_test_cor.append(first_pred)

  batch_new_cor = np.append(batch_new_cor[:,1:,:], [[first_pred]], axis = 1)

1/1 [==============================] - 1s 1s/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 26ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
In [146]:
#Changing to array
predict_test_cor = np.array(predict_test_cor)
#Unscaling values
predictions_cor = sc2.inverse_transform(predict_test_cor)

Results for Corn¶

In [147]:
#Plot comparison between predicted and actual values
plt.plot(corn_LSTM_test,color = 'red', label = 'Actual Values')
plt.plot(predictions_cor, color = 'blue', label = 'Predicted Values')
plt.title('Predicted VS Actual Values')
plt.xlabel('Time (M)')
plt.ylabel('Corn Price')
plt.legend()
plt.show()
In [148]:
#Performace metrics
RMSE_corn = math.sqrt(mean_squared_error(corn_LSTM_test, predictions_cor))
In [149]:
RMSE_corn
Out[149]:
1.7238802156265998
In [150]:
mae(corn_LSTM_test, predictions_cor)
Out[150]:
1.5645409348676236

LSTM Coffee ¶

Individual Pre-Processing for Coffee¶

In [151]:
coffee_LSTM = coffee.resample('M').mean()
In [152]:
x4 = len(coffee_LSTM)
size4 = int(x4 * 0.8)
index4 = x4 - size4
coffee_LSTM_train = coffee_LSTM.iloc[:size4,].values
coffee_LSTM_test = coffee_LSTM.iloc[size4:,].values
In [153]:
coffee_LSTM_train_scaled = sc2.fit_transform(coffee_LSTM_train)
coffee_LSTM_test_scaled = sc2.fit_transform(coffee_LSTM_test)
In [154]:
#Lists for X and Y train
x_train_cof = []
y_train_cof = []
#Window size
WS = 10
#Creating arrays for model
for i in range(WS, len(coffee_LSTM_train_scaled)):
  x_train_cof.append(coffee_LSTM_train_scaled[i-WS:i,0:1])
  y_train_cof.append(coffee_LSTM_train_scaled[i,0])

x_train_cof, y_train_cof =np.array(x_train_cof), np.array(y_train_cof)

Gridsearch for Coffee hyperparameters

In [155]:
def create_model_cof(units=50, activation='relu', optimizer='adam'):
    model = Sequential()
    model.add(LSTM(units=units, activation=activation, input_shape=(x_train_cof.shape[1], x_train_cof.shape[2])))
    model.add(Dense(1))
    model.compile(loss='mean_squared_error', optimizer=optimizer)
    return model
In [156]:
model_cof_tune = KerasRegressor(build_fn=create_model_cof, verbose=0)
In [157]:
grid_cof = GridSearchCV(estimator=model_cof_tune, param_grid=param_grid, scoring='neg_mean_squared_error', cv=3)
In [158]:
grid_result_cof = grid_cof.fit(x_train_cof, y_train_cof)
In [159]:
print("Best: %f using %s" % (grid_result_cof.best_score_, grid_result_cof.best_params_))

Best: -0.040449 using {'activation': 'relu', 'optimizer': 'adam', 'units': 300}

Developing LSTM model for Coffee¶

In [160]:
#First Layer
model_cof = Sequential()
model_cof.add(LSTM(units = 300, activation = 'relu', return_sequences = True, input_shape = (x_train_cof.shape[1],1)))

#Dropout to avoid over fitting
model_cof.add(Dropout(0.2))

#Next layer no input - first hidden layer
model_cof.add(LSTM(units = 300, return_sequences = True))
model_cof.add(Dropout(0.2))

#Second hidden layer
model_cof.add(LSTM(units = 300, return_sequences = True))
model_cof.add(Dropout(0.2))

#Last layer- no return
model_cof.add(LSTM(units = 300))
model_cof.add(Dropout(0.2))

#Output
model_cof.add(Dense(units = 1))

#Optimizer 
model_cof.compile(optimizer = 'rmsprop', loss = 'mean_squared_error')

#Fitting model
model_cof.fit(x_train_cof,y_train_cof, epochs = 50, batch_size = 5)

Epoch 1/50
94/94 [==============================] - 11s 51ms/step - loss: 0.0993
Epoch 2/50
94/94 [==============================] - 4s 47ms/step - loss: 0.0552
Epoch 3/50
94/94 [==============================] - 4s 46ms/step - loss: 0.0390
Epoch 4/50
94/94 [==============================] - 4s 46ms/step - loss: 0.0310
Epoch 5/50
94/94 [==============================] - 5s 48ms/step - loss: 0.0268
Epoch 6/50
94/94 [==============================] - 5s 48ms/step - loss: 0.0234
Epoch 7/50
94/94 [==============================] - 5s 50ms/step - loss: 0.0191
Epoch 8/50
94/94 [==============================] - 4s 46ms/step - loss: 0.0186
Epoch 9/50
94/94 [==============================] - 4s 47ms/step - loss: 0.0183
Epoch 10/50
94/94 [==============================] - 5s 49ms/step - loss: 0.0176
Epoch 11/50
94/94 [==============================] - 4s 47ms/step - loss: 0.0167
Epoch 12/50
94/94 [==============================] - 5s 48ms/step - loss: 0.0165
Epoch 13/50
94/94 [==============================] - 5s 49ms/step - loss: 0.0157
Epoch 14/50
94/94 [==============================] - 5s 53ms/step - loss: 0.0139
Epoch 15/50
94/94 [==============================] - 4s 48ms/step - loss: 0.0144
Epoch 16/50
94/94 [==============================] - 4s 46ms/step - loss: 0.0148
Epoch 17/50
94/94 [==============================] - 5s 49ms/step - loss: 0.0143
Epoch 18/50
94/94 [==============================] - 4s 47ms/step - loss: 0.0147
Epoch 19/50
94/94 [==============================] - 4s 46ms/step - loss: 0.0141
Epoch 20/50
94/94 [==============================] - 4s 47ms/step - loss: 0.0154
Epoch 21/50
94/94 [==============================] - 4s 45ms/step - loss: 0.0144
Epoch 22/50
94/94 [==============================] - 4s 46ms/step - loss: 0.0138
Epoch 23/50
94/94 [==============================] - 5s 48ms/step - loss: 0.0147
Epoch 24/50
94/94 [==============================] - 4s 47ms/step - loss: 0.0122
Epoch 25/50
94/94 [==============================] - 4s 48ms/step - loss: 0.0143
Epoch 26/50
94/94 [==============================] - 4s 47ms/step - loss: 0.0131
Epoch 27/50
94/94 [==============================] - 5s 49ms/step - loss: 0.0137
Epoch 28/50
94/94 [==============================] - 4s 47ms/step - loss: 0.0133
Epoch 29/50
94/94 [==============================] - 4s 46ms/step - loss: 0.0134
Epoch 30/50
94/94 [==============================] - 5s 48ms/step - loss: 0.0134
Epoch 31/50
94/94 [==============================] - 4s 47ms/step - loss: 0.0124
Epoch 32/50
94/94 [==============================] - 5s 48ms/step - loss: 0.0121
Epoch 33/50
94/94 [==============================] - 4s 47ms/step - loss: 0.0122
Epoch 34/50
94/94 [==============================] - 5s 48ms/step - loss: 0.0115
Epoch 35/50
94/94 [==============================] - 4s 47ms/step - loss: 0.0128
Epoch 36/50
94/94 [==============================] - 4s 46ms/step - loss: 0.0126
Epoch 37/50
94/94 [==============================] - 5s 49ms/step - loss: 0.0121
Epoch 38/50
94/94 [==============================] - 4s 45ms/step - loss: 0.0113
Epoch 39/50
94/94 [==============================] - 4s 47ms/step - loss: 0.0116
Epoch 40/50
94/94 [==============================] - 5s 50ms/step - loss: 0.0118
Epoch 41/50
94/94 [==============================] - 5s 48ms/step - loss: 0.0112
Epoch 42/50
94/94 [==============================] - 4s 47ms/step - loss: 0.0124
Epoch 43/50
94/94 [==============================] - 4s 46ms/step - loss: 0.0130
Epoch 44/50
94/94 [==============================] - 5s 49ms/step - loss: 0.0117
Epoch 45/50
94/94 [==============================] - 5s 48ms/step - loss: 0.0112
Epoch 46/50
94/94 [==============================] - 4s 46ms/step - loss: 0.0112
Epoch 47/50
94/94 [==============================] - 5s 49ms/step - loss: 0.0109
Epoch 48/50
94/94 [==============================] - 4s 46ms/step - loss: 0.0132
Epoch 49/50
94/94 [==============================] - 4s 47ms/step - loss: 0.0117
Epoch 50/50
94/94 [==============================] - 5s 49ms/step - loss: 0.0128
Out[160]:
<keras.callbacks.History at 0x7f8cfc589cd0>
In [161]:
#Evaluating the number of epochs
plt.plot(range(len(model_cof.history.history['loss'])),model_cof.history.history['loss'])
plt.xlabel('Epoch Number')
plt.ylabel('Loss')
plt.show()

Predictions for Coffee¶

In [162]:
#Prediction setup
predict_test_cof = [] 
batch_one_cof = coffee_LSTM_train_scaled[-WS:]
batch_new_cof = batch_one_cof.reshape((1,WS,1))

#Set number of predictions in range.
for i in range(len(coffee_LSTM_test_scaled)):

  first_pred = model_cof.predict(batch_new_cof)[0]

  predict_test_cof.append(first_pred)

  batch_new_cof = np.append(batch_new_cof[:,1:,:], [[first_pred]], axis = 1)

1/1 [==============================] - 1s 1s/step
1/1 [==============================] - 0s 33ms/step
1/1 [==============================] - 0s 33ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 33ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 32ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 32ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 32ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 33ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 32ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 30ms/step
In [163]:
#Changing to array
predict_test_cof = np.array(predict_test_cof)
#Unscaling values
predictions_cof = sc2.inverse_transform(predict_test_cof)

Results for Coffee¶

In [164]:
#Plot comparison between predicted and actual values
plt.plot(coffee_LSTM_test,color = 'red', label = 'Actual Values')
plt.plot(predictions_cof, color = 'blue', label = 'Predicted Values')
plt.title('Predicted VS Actual Values')
plt.xlabel('Time (M)')
plt.ylabel('Coffee Price')
plt.legend()
plt.show()
In [165]:
coffee_LSTM_test = coffee_LSTM_test[~np.isnan(coffee_LSTM_test)]
#performace metrics
RMSE_coffee = math.sqrt(mean_squared_error(coffee_LSTM_test, predictions_cof[:119]))
In [166]:
RMSE_coffee
Out[166]:
1.4051282061944668
In [167]:
mae(coffee_LSTM_test, predictions_cof[:119])
Out[167]:
1.3086615853416776

LSTM Soybean ¶

Individual Pre-Processing for Soybean¶

In [168]:
soybean_LSTM = soybean.resample('M').mean()
In [169]:
x5 = len(soybean_LSTM)
size5 = int(x5 * 0.8)
index5 = x5 - size5
soybean_LSTM_train = soybean_LSTM.iloc[:size5,].values
soybean_LSTM_test = soybean_LSTM.iloc[size5:,].values
In [170]:
soybean_LSTM_train_scaled = sc2.fit_transform(soybean_LSTM_train)
soybean_LSTM_test_scaled = sc2.fit_transform(soybean_LSTM_test)
In [171]:
#Lists for X an Y train
x_train_sb = []
y_train_sb = []
#Window size
WS = 10
#creating arrays for model
for i in range(WS, len(soybean_LSTM_train_scaled)):
  x_train_sb.append(soybean_LSTM_train_scaled[i-WS:i,0:1])
  y_train_sb.append(soybean_LSTM_train_scaled[i,0])

x_train_sb, y_train_sb =np.array(x_train_sb), np.array(y_train_sb)

Gridsearch for Soybean hyperparameters

In [172]:
def create_model_sb(units=50, activation='relu', optimizer='adam'):
    model = Sequential()
    model.add(LSTM(units=units, activation=activation, input_shape=(x_train_sb.shape[1], x_train_sb.shape[2])))
    model.add(Dense(1))
    model.compile(loss='mean_squared_error', optimizer=optimizer)
    return model
In [173]:
model_sb_tune = KerasRegressor(build_fn=create_model_sb, verbose=0)
In [174]:
grid_sb = GridSearchCV(estimator=model_sb_tune, param_grid=param_grid, scoring='neg_mean_squared_error', cv=3)
In [175]:
grid_result_sb = grid_sb.fit(x_train_sb, y_train_sb)
In [176]:
print("Best: %f using %s" % (grid_result_sb.best_score_, grid_result_sb.best_params_))

Best: -0.036753 using {'activation': 'tanh', 'optimizer': 'adam', 'units': 200}

Developing LSTM Model for Soybean¶

In [177]:
#First Layer
model_sb = Sequential()
model_sb.add(LSTM(units = 300, activation = 'tanh', return_sequences = True, input_shape = (x_train_sb.shape[1],1)))

#Dropout to avoid over fitting
model_sb.add(Dropout(0.2))

#Next layer no input - first hidden layer
model_sb.add(LSTM(units = 300, return_sequences = True))
model_sb.add(Dropout(0.2))

#Second hidden layer
model_sb.add(LSTM(units =300, return_sequences = True))
model_sb.add(Dropout(0.2))

#Last layer- no return
model_sb.add(LSTM(units = 300))
model_sb.add(Dropout(0.2))

#Output
model_sb.add(Dense(units = 1))

#Optimizer
model_sb.compile(optimizer = 'rmsprop', loss = 'mean_squared_error')

#Fitting model
model_sb.fit(x_train_sb,y_train_sb, epochs = 50, batch_size = 5)

Epoch 1/50
102/102 [==============================] - 10s 47ms/step - loss: 0.0731
Epoch 2/50
102/102 [==============================] - 4s 44ms/step - loss: 0.0472
Epoch 3/50
102/102 [==============================] - 4s 42ms/step - loss: 0.0386
Epoch 4/50
102/102 [==============================] - 5s 46ms/step - loss: 0.0261
Epoch 5/50
102/102 [==============================] - 4s 44ms/step - loss: 0.0215
Epoch 6/50
102/102 [==============================] - 4s 44ms/step - loss: 0.0175
Epoch 7/50
102/102 [==============================] - 5s 45ms/step - loss: 0.0147
Epoch 8/50
102/102 [==============================] - 4s 43ms/step - loss: 0.0139
Epoch 9/50
102/102 [==============================] - 5s 44ms/step - loss: 0.0108
Epoch 10/50
102/102 [==============================] - 5s 44ms/step - loss: 0.0108
Epoch 11/50
102/102 [==============================] - 4s 44ms/step - loss: 0.0121
Epoch 12/50
102/102 [==============================] - 4s 42ms/step - loss: 0.0116
Epoch 13/50
102/102 [==============================] - 4s 44ms/step - loss: 0.0108
Epoch 14/50
102/102 [==============================] - 5s 46ms/step - loss: 0.0125
Epoch 15/50
102/102 [==============================] - 5s 45ms/step - loss: 0.0109
Epoch 16/50
102/102 [==============================] - 5s 44ms/step - loss: 0.0105
Epoch 17/50
102/102 [==============================] - 5s 46ms/step - loss: 0.0100
Epoch 18/50
102/102 [==============================] - 5s 45ms/step - loss: 0.0099
Epoch 19/50
102/102 [==============================] - 5s 45ms/step - loss: 0.0102
Epoch 20/50
102/102 [==============================] - 5s 46ms/step - loss: 0.0108
Epoch 21/50
102/102 [==============================] - 5s 46ms/step - loss: 0.0097
Epoch 22/50
102/102 [==============================] - 5s 48ms/step - loss: 0.0100
Epoch 23/50
102/102 [==============================] - 5s 45ms/step - loss: 0.0104
Epoch 24/50
102/102 [==============================] - 4s 43ms/step - loss: 0.0105
Epoch 25/50
102/102 [==============================] - 4s 42ms/step - loss: 0.0110
Epoch 26/50
102/102 [==============================] - 5s 44ms/step - loss: 0.0103
Epoch 27/50
102/102 [==============================] - 5s 46ms/step - loss: 0.0099
Epoch 28/50
102/102 [==============================] - 5s 45ms/step - loss: 0.0093
Epoch 29/50
102/102 [==============================] - 5s 45ms/step - loss: 0.0097
Epoch 30/50
102/102 [==============================] - 5s 44ms/step - loss: 0.0099
Epoch 31/50
102/102 [==============================] - 4s 44ms/step - loss: 0.0089
Epoch 32/50
102/102 [==============================] - 4s 43ms/step - loss: 0.0085
Epoch 33/50
102/102 [==============================] - 4s 43ms/step - loss: 0.0100
Epoch 34/50
102/102 [==============================] - 4s 44ms/step - loss: 0.0095
Epoch 35/50
102/102 [==============================] - 4s 43ms/step - loss: 0.0095
Epoch 36/50
102/102 [==============================] - 5s 45ms/step - loss: 0.0094
Epoch 37/50
102/102 [==============================] - 5s 45ms/step - loss: 0.0090
Epoch 38/50
102/102 [==============================] - 4s 43ms/step - loss: 0.0082
Epoch 39/50
102/102 [==============================] - 4s 41ms/step - loss: 0.0102
Epoch 40/50
102/102 [==============================] - 5s 48ms/step - loss: 0.0094
Epoch 41/50
102/102 [==============================] - 5s 47ms/step - loss: 0.0097
Epoch 42/50
102/102 [==============================] - 5s 50ms/step - loss: 0.0093
Epoch 43/50
102/102 [==============================] - 5s 45ms/step - loss: 0.0102
Epoch 44/50
102/102 [==============================] - 5s 46ms/step - loss: 0.0090
Epoch 45/50
102/102 [==============================] - 5s 45ms/step - loss: 0.0087
Epoch 46/50
102/102 [==============================] - 5s 46ms/step - loss: 0.0080
Epoch 47/50
102/102 [==============================] - 5s 44ms/step - loss: 0.0086
Epoch 48/50
102/102 [==============================] - 4s 43ms/step - loss: 0.0085
Epoch 49/50
102/102 [==============================] - 4s 44ms/step - loss: 0.0089
Epoch 50/50
102/102 [==============================] - 5s 46ms/step - loss: 0.0095
Out[177]:
<keras.callbacks.History at 0x7f8dcc4cf370>
In [178]:
#Evaluating the number of epochs
plt.plot(range(len(model_sb.history.history['loss'])),model_sb.history.history['loss'])
plt.xlabel('Epoch Number')
plt.ylabel('Loss')
plt.show()

Predictions for Soybean¶

In [179]:
#Prediction steup
predict_test_sb = [] 
batch_one_sb = soybean_LSTM_train_scaled[-WS:]
batch_new_sb = batch_one_sb.reshape((1,WS,1))

#Set number of predictions in range.
for i in range(len(soybean_LSTM_test)):

  first_pred = model_sb.predict(batch_new_sb)[0]

  predict_test_sb.append(first_pred)

  batch_new_sb = np.append(batch_new_sb[:,1:,:], [[first_pred]], axis = 1)

1/1 [==============================] - 1s 1s/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 39ms/step
1/1 [==============================] - 0s 37ms/step
1/1 [==============================] - 0s 37ms/step
1/1 [==============================] - 0s 38ms/step
1/1 [==============================] - 0s 37ms/step
1/1 [==============================] - 0s 37ms/step
1/1 [==============================] - 0s 38ms/step
1/1 [==============================] - 0s 39ms/step
1/1 [==============================] - 0s 36ms/step
1/1 [==============================] - 0s 37ms/step
1/1 [==============================] - 0s 38ms/step
1/1 [==============================] - 0s 38ms/step
1/1 [==============================] - 0s 38ms/step
1/1 [==============================] - 0s 39ms/step
1/1 [==============================] - 0s 36ms/step
1/1 [==============================] - 0s 37ms/step
1/1 [==============================] - 0s 36ms/step
1/1 [==============================] - 0s 36ms/step
1/1 [==============================] - 0s 36ms/step
1/1 [==============================] - 0s 36ms/step
1/1 [==============================] - 0s 37ms/step
1/1 [==============================] - 0s 35ms/step
1/1 [==============================] - 0s 38ms/step
1/1 [==============================] - 0s 37ms/step
1/1 [==============================] - 0s 38ms/step
1/1 [==============================] - 0s 38ms/step
1/1 [==============================] - 0s 38ms/step
1/1 [==============================] - 0s 38ms/step
1/1 [==============================] - 0s 37ms/step
1/1 [==============================] - 0s 39ms/step
1/1 [==============================] - 0s 36ms/step
1/1 [==============================] - 0s 38ms/step
1/1 [==============================] - 0s 36ms/step
1/1 [==============================] - 0s 37ms/step
1/1 [==============================] - 0s 36ms/step
1/1 [==============================] - 0s 37ms/step
1/1 [==============================] - 0s 37ms/step
1/1 [==============================] - 0s 38ms/step
1/1 [==============================] - 0s 38ms/step
1/1 [==============================] - 0s 38ms/step
1/1 [==============================] - 0s 39ms/step
1/1 [==============================] - 0s 38ms/step
1/1 [==============================] - 0s 40ms/step
1/1 [==============================] - 0s 41ms/step
1/1 [==============================] - 0s 36ms/step
1/1 [==============================] - 0s 39ms/step
1/1 [==============================] - 0s 39ms/step
1/1 [==============================] - 0s 40ms/step
1/1 [==============================] - 0s 43ms/step
1/1 [==============================] - 0s 43ms/step
1/1 [==============================] - 0s 41ms/step
1/1 [==============================] - 0s 40ms/step
1/1 [==============================] - 0s 40ms/step
1/1 [==============================] - 0s 38ms/step
1/1 [==============================] - 0s 41ms/step
1/1 [==============================] - 0s 42ms/step
1/1 [==============================] - 0s 41ms/step
1/1 [==============================] - 0s 39ms/step
1/1 [==============================] - 0s 39ms/step
1/1 [==============================] - 0s 40ms/step
1/1 [==============================] - 0s 37ms/step
1/1 [==============================] - 0s 36ms/step
1/1 [==============================] - 0s 41ms/step
1/1 [==============================] - 0s 41ms/step
1/1 [==============================] - 0s 38ms/step
1/1 [==============================] - 0s 38ms/step
1/1 [==============================] - 0s 35ms/step
1/1 [==============================] - 0s 34ms/step
1/1 [==============================] - 0s 35ms/step
1/1 [==============================] - 0s 34ms/step
1/1 [==============================] - 0s 38ms/step
1/1 [==============================] - 0s 38ms/step
1/1 [==============================] - 0s 35ms/step
1/1 [==============================] - 0s 34ms/step
1/1 [==============================] - 0s 34ms/step
1/1 [==============================] - 0s 34ms/step
1/1 [==============================] - 0s 34ms/step
1/1 [==============================] - 0s 33ms/step
1/1 [==============================] - 0s 33ms/step
1/1 [==============================] - 0s 39ms/step
1/1 [==============================] - 0s 37ms/step
1/1 [==============================] - 0s 34ms/step
1/1 [==============================] - 0s 35ms/step
1/1 [==============================] - 0s 35ms/step
1/1 [==============================] - 0s 33ms/step
1/1 [==============================] - 0s 32ms/step
1/1 [==============================] - 0s 33ms/step
1/1 [==============================] - 0s 33ms/step
In [180]:
#Changing to array
predict_test_sb = np.array(predict_test_sb)
#Unscaling values
predictions_sb = sc2.inverse_transform(predict_test_sb)

Results for Soybean¶

In [181]:
#Plot comparison between predicted and actual values
plt.plot(soybean_LSTM_test,color = 'red', label = 'Actual Values')
plt.plot(predictions_sb, color = 'blue', label = 'Predicted Values')
plt.title('Predicted VS Actual Values')
plt.xlabel('Time (M)')
plt.ylabel('Soybean Price')
plt.legend()
plt.show()
In [182]:
#Performace metrics
RMSE_soybean = math.sqrt(mean_squared_error(soybean_LSTM_test, predictions_sb))
In [183]:
RMSE_soybean
Out[183]:
2.7415217559025096
In [184]:
mae(soybean_LSTM_test, predictions_sb)
Out[184]:
1.8143227754994755

LSTM Soybean Oil ¶

Individual Pre-Processing for Soybean Oil¶

In [185]:
soybean_oil_LSTM = soybean_oil.resample('M').mean()
In [186]:
x6 = len(soybean_oil_LSTM)
size6 = int(x6 * 0.8)
index6 = x6 - size6
soybean_oil_LSTM_train = soybean_oil_LSTM.iloc[:size6,].values
soybean_oil_LSTM_test = soybean_oil_LSTM.iloc[size6:,].values
In [187]:
soybean_oil_LSTM_train_scaled = sc2.fit_transform(soybean_oil_LSTM_train)
soybean_oil_LSTM_test_scaled = sc2.fit_transform(soybean_oil_LSTM_test)
In [188]:
#Lists for X an Y train
x_train_sbo = []
y_train_sbo = []
#Window size
WS = 12
#Creating arrays for model
for i in range(WS, len(soybean_oil_LSTM_train_scaled)):
  x_train_sbo.append(soybean_oil_LSTM_train_scaled[i-WS:i,0:1])
  y_train_sbo.append(soybean_oil_LSTM_train_scaled[i,0])

x_train_sbo, y_train_sbo =np.array(x_train_sbo), np.array(y_train_sbo)

Gridsearch for Soybean Oil Hyperparameters

In [189]:
def create_model_sbo(units=50, activation='relu', optimizer='adam'):
    model = Sequential()
    model.add(LSTM(units=units, activation=activation, input_shape=(x_train_sbo.shape[1], x_train_sbo.shape[2])))
    model.add(Dense(1))
    model.compile(loss='mean_squared_error', optimizer=optimizer)
    return model
In [190]:
model_sbo_tune = KerasRegressor(build_fn= create_model_sbo, verbose=0)
In [191]:
grid_sbo = GridSearchCV(estimator=model_sbo_tune, param_grid=param_grid, scoring='neg_mean_squared_error', cv=3)
In [192]:
grid_result_sbo = grid_sbo.fit(x_train_sbo, y_train_sbo)
In [193]:
print("Best: %f using %s" % (grid_result_sbo.best_score_, grid_result_sbo.best_params_))

Best: -0.025069 using {'activation': 'tanh', 'optimizer': 'rmsprop', 'units': 60}

Developing LSTM Model for Soybean Oil¶

In [ ]:
#First layer
model_sbo = Sequential()
model_sbo.add(LSTM(units = 200, activation = 'tanh', return_sequences = True, input_shape = (x_train_sbo.shape[1],1)))

#Dropout to avoid over fitting
model_sbo.add(Dropout(0.2))

#Next layer no input - first hidden layer
model_sbo.add(LSTM(units = 200, return_sequences = True))
model_sbo.add(Dropout(0.2))

#Second hidden layer
model_sbo.add(LSTM(units = 200, return_sequences = True))
model_sbo.add(Dropout(0.2))

#Last layer- no return
model_sbo.add(LSTM(units = 200))
model_sbo.add(Dropout(0.2))

#Output
model_sbo.add(Dense(units = 1))

#Optimizer 'rmsprop' - chosen through Gridsearch
model_sbo.compile(optimizer = 'rmsprop', loss = 'mean_squared_error')

#Fitting model
model_sbo.fit(x_train_sbo,y_train_sbo, epochs = 50, batch_size = 5)

Epoch 1/50
116/116 [==============================] - 11s 46ms/step - loss: 0.0728
Epoch 2/50
116/116 [==============================] - 5s 46ms/step - loss: 0.0437
Epoch 3/50
116/116 [==============================] - 5s 46ms/step - loss: 0.0327
Epoch 4/50
116/116 [==============================] - 5s 47ms/step - loss: 0.0237
Epoch 5/50
116/116 [==============================] - 5s 47ms/step - loss: 0.0206
Epoch 6/50
116/116 [==============================] - 5s 46ms/step - loss: 0.0157
Epoch 7/50
116/116 [==============================] - 5s 47ms/step - loss: 0.0146
Epoch 8/50
116/116 [==============================] - 5s 46ms/step - loss: 0.0118
Epoch 9/50
116/116 [==============================] - 5s 46ms/step - loss: 0.0111
Epoch 10/50
116/116 [==============================] - 6s 47ms/step - loss: 0.0099
Epoch 11/50
116/116 [==============================] - 5s 46ms/step - loss: 0.0097
Epoch 12/50
116/116 [==============================] - 5s 47ms/step - loss: 0.0091
Epoch 13/50
116/116 [==============================] - 5s 46ms/step - loss: 0.0088
Epoch 14/50
116/116 [==============================] - 5s 45ms/step - loss: 0.0083
Epoch 15/50
116/116 [==============================] - 6s 48ms/step - loss: 0.0067
Epoch 16/50
116/116 [==============================] - 5s 46ms/step - loss: 0.0086
Epoch 17/50
116/116 [==============================] - 5s 47ms/step - loss: 0.0088
Epoch 18/50
116/116 [==============================] - 6s 48ms/step - loss: 0.0077
Epoch 19/50
116/116 [==============================] - 5s 47ms/step - loss: 0.0077
Epoch 20/50
116/116 [==============================] - 5s 46ms/step - loss: 0.0075
Epoch 21/50
116/116 [==============================] - 5s 45ms/step - loss: 0.0069
Epoch 22/50
116/116 [==============================] - 5s 45ms/step - loss: 0.0069
Epoch 23/50
116/116 [==============================] - 5s 45ms/step - loss: 0.0075
Epoch 24/50
116/116 [==============================] - 5s 45ms/step - loss: 0.0073
Epoch 25/50
116/116 [==============================] - 5s 45ms/step - loss: 0.0067
Epoch 26/50
116/116 [==============================] - 5s 45ms/step - loss: 0.0069
Epoch 27/50
116/116 [==============================] - 5s 45ms/step - loss: 0.0065
Epoch 28/50
116/116 [==============================] - 5s 46ms/step - loss: 0.0075
Epoch 29/50
116/116 [==============================] - 5s 44ms/step - loss: 0.0069
Epoch 30/50
116/116 [==============================] - 5s 43ms/step - loss: 0.0064
Epoch 31/50
116/116 [==============================] - 5s 45ms/step - loss: 0.0075
Epoch 32/50
116/116 [==============================] - 5s 43ms/step - loss: 0.0069
Epoch 33/50
116/116 [==============================] - 5s 45ms/step - loss: 0.0070
Epoch 34/50
116/116 [==============================] - 5s 46ms/step - loss: 0.0070
Epoch 35/50
116/116 [==============================] - 5s 44ms/step - loss: 0.0077
Epoch 36/50
116/116 [==============================] - 5s 45ms/step - loss: 0.0062
Epoch 37/50
116/116 [==============================] - 5s 44ms/step - loss: 0.0070
Epoch 38/50
116/116 [==============================] - 5s 44ms/step - loss: 0.0067
Epoch 39/50
116/116 [==============================] - 5s 45ms/step - loss: 0.0071
Epoch 40/50
116/116 [==============================] - 5s 44ms/step - loss: 0.0061
Epoch 41/50
116/116 [==============================] - 5s 44ms/step - loss: 0.0068
Epoch 42/50
116/116 [==============================] - 5s 44ms/step - loss: 0.0055
Epoch 43/50
116/116 [==============================] - 5s 43ms/step - loss: 0.0069
Epoch 44/50
116/116 [==============================] - 5s 44ms/step - loss: 0.0059
Epoch 45/50
116/116 [==============================] - 5s 45ms/step - loss: 0.0061
Epoch 46/50
116/116 [==============================] - 5s 43ms/step - loss: 0.0070
Epoch 47/50
116/116 [==============================] - 5s 44ms/step - loss: 0.0065
Epoch 48/50
116/116 [==============================] - 5s 44ms/step - loss: 0.0062
Epoch 49/50
116/116 [==============================] - 5s 44ms/step - loss: 0.0063
Epoch 50/50
116/116 [==============================] - 5s 45ms/step - loss: 0.0063
Out[ ]:
<keras.callbacks.History at 0x7f8cfdac51f0>
In [ ]:
#Evlauating the number of epochs
plt.plot(range(len(model_sbo.history.history['loss'])),model_sbo.history.history['loss'])
plt.xlabel('Epoch Number')
plt.ylabel('Loss')
plt.show()

Predictions for Soybean Oil¶

In [ ]:
#Prediction setup
predict_test_sbo = [] 
batch_one_sbo = soybean_oil_LSTM_train_scaled[-WS:]
batch_new_sbo = batch_one_sbo.reshape((1,WS,1))

#Set number of predictions in range
for i in range(149):

  first_pred = model_sbo.predict(batch_new_sbo)[0]

  predict_test_sbo.append(first_pred)

  batch_new_sbo = np.append(batch_new_sbo[:,1:,:],[[first_pred]], axis = 1)

1/1 [==============================] - 1s 1s/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 26ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 26ms/step
1/1 [==============================] - 0s 26ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 31ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 29ms/step
1/1 [==============================] - 0s 30ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 27ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
1/1 [==============================] - 0s 28ms/step
In [ ]:
#Changing to array
predict_test_sbo = np.array(predict_test_sbo)
#Unscaling values
predictions_sbo = sc2.inverse_transform(predict_test_sbo)

Results for Soybean Oil¶

In [ ]:
#Plot comparison between predicted and actual values
plt.plot(soybean_oil_LSTM_test,color = 'red', label = 'Actual Values')
plt.plot(predictions_sbo, color = 'blue', label = 'Predicted Values')
plt.title('Predicted VS Actual Values')
plt.xlabel('Time (M)')
plt.ylabel('Soybean Price')
plt.legend()
plt.show()
In [ ]:
#Performace metrics
RMSE_soybean_oil = math.sqrt(mean_squared_error(soybean_oil_LSTM_test, predictions_sbo))
In [ ]:
RMSE_soybean_oil
Out[ ]:
0.2689092902261356
In [ ]:
mae(soybean_oil_LSTM_test, predictions_sbo)
Out[ ]:
0.24252537532362348

Results¶

In [ ]:
#Aluminum
print('RMSE Aluminum - ARIMA')
print(RMSE_ARIMA_aluminum)

RMSE Aluminum - ARIMA
505.2406577643067
In [ ]:
print('RMSE Aluminum - LSTM')
print(Aluminum_RMSE)

RMSE Aluminum - LSTM
502.53211929570585
In [ ]:
#Sugar
print('RMSE Sugar - ARIMA')
print(RMSE_ARIMA_sugar)

RMSE Sugar - ARIMA
0.02885330600594616
In [ ]:
print('RMSE Sugar- LSTM')
print(RMSE_sugar)

RMSE Sugar- LSTM
0.042438750327986836
In [ ]:
#Cotton
print('RMSE Cotton - ARIMA')
print(RMSE_ARIMA_cotton)

RMSE Cotton - ARIMA
0.257799108563674
In [ ]:
print('RMSE Cotton - LSTM')
print(RMSE_cotton)

RMSE Cotton - LSTM
0.7173651109058645
In [ ]:
#Corn
print('RMSE Corn - ARIMA')
print(RMSE_ARIMA_corn)

RMSE Corn - ARIMA
1.8382735266248653
In [ ]:
print('RMSE Corn - LSTM')
print(RMSE_corn)

RMSE Corn - LSTM
1.7238802156265998
In [ ]:
#Coffee
print('RMSE Coffee - ARIMA')
print(RMSE_ARIMA_coffee)

RMSE Coffee - ARIMA
0.5086124942858565
In [ ]:
print('RMSE Coffee - LSTM')
print(RMSE_coffee)

RMSE Coffee - LSTM
1.4051282061944668
In [ ]:
#Soybean
print('RMSE Soybean - ARIMA')
print(RMSE_ARIMA_soybean)

RMSE Soybean - ARIMA
3.367017508721487
In [ ]:
print('RMSE Soybean - LSTM')
print(RMSE_soybean)

RMSE Soybean - LSTM
2.7415217559025096
In [ ]:
#Soybean oil
print('RMSE Soybean Oil - ARIMA')
print(RMSE_ARIMA_soybean_oil)

RMSE Soybean Oil - ARIMA
0.21916021303429323
In [ ]:
print('RMSE Soybean Oil - LSTM')
print(RMSE_soybean_oil)

RMSE Soybean Oil - LSTM
0.2689092902261356

The ARIMA model outperformed the LSTM model for every commodity. The LSTM model may depend more on seasonality and our data is not highly seasonal. The cyclical nature does not seem to benefit this recurrent neural network model. Of the two models, I would work with the ARIMA because of the better fit overall and the relative consistency of results produced. I had trouble getting consistent results from runtime to runtime with the LSTM model.

Export¶

In [ ]:
!cp "/content/drive/MyDrive/Colab Notebooks/EDA/Final_Individual_Notebook_Stoker.ipynb" ./

!jupyter nbconvert --to html "Final_Individual_Notebook_Stoker.ipynb"

cp: cannot stat '/content/drive/MyDrive/Colab Notebooks/EDA/Final_Individual_Notebook_Stoker.ipynb': No such file or directory
[NbConvertApp] WARNING | pattern 'Final_Individual_Notebook_Stoker.ipynb' matched no files
This application is used to convert notebook files (*.ipynb)
        to various other formats.

        WARNING: THE COMMANDLINE INTERFACE MAY CHANGE IN FUTURE RELEASES.

Options
=======
The options below are convenience aliases to configurable class-options,
as listed in the "Equivalent to" description-line of the aliases.
To see all configurable class-options for some <cmd>, use:
    <cmd> --help-all

--debug
    set log level to logging.DEBUG (maximize logging output)
    Equivalent to: [--Application.log_level=10]
--show-config
    Show the application's configuration (human-readable format)
    Equivalent to: [--Application.show_config=True]
--show-config-json
    Show the application's configuration (json format)
    Equivalent to: [--Application.show_config_json=True]
--generate-config
    generate default config file
    Equivalent to: [--JupyterApp.generate_config=True]
-y
    Answer yes to any questions instead of prompting.
    Equivalent to: [--JupyterApp.answer_yes=True]
--execute
    Execute the notebook prior to export.
    Equivalent to: [--ExecutePreprocessor.enabled=True]
--allow-errors
    Continue notebook execution even if one of the cells throws an error and include the error message in the cell output (the default behaviour is to abort conversion). This flag is only relevant if '--execute' was specified, too.
    Equivalent to: [--ExecutePreprocessor.allow_errors=True]
--stdin
    read a single notebook file from stdin. Write the resulting notebook with default basename 'notebook.*'
    Equivalent to: [--NbConvertApp.from_stdin=True]
--stdout
    Write notebook output to stdout instead of files.
    Equivalent to: [--NbConvertApp.writer_class=StdoutWriter]
--inplace
    Run nbconvert in place, overwriting the existing notebook (only
            relevant when converting to notebook format)
    Equivalent to: [--NbConvertApp.use_output_suffix=False --NbConvertApp.export_format=notebook --FilesWriter.build_directory=]
--clear-output
    Clear output of current file and save in place,
            overwriting the existing notebook.
    Equivalent to: [--NbConvertApp.use_output_suffix=False --NbConvertApp.export_format=notebook --FilesWriter.build_directory= --ClearOutputPreprocessor.enabled=True]
--no-prompt
    Exclude input and output prompts from converted document.
    Equivalent to: [--TemplateExporter.exclude_input_prompt=True --TemplateExporter.exclude_output_prompt=True]
--no-input
    Exclude input cells and output prompts from converted document.
            This mode is ideal for generating code-free reports.
    Equivalent to: [--TemplateExporter.exclude_output_prompt=True --TemplateExporter.exclude_input=True --TemplateExporter.exclude_input_prompt=True]
--allow-chromium-download
    Whether to allow downloading chromium if no suitable version is found on the system.
    Equivalent to: [--WebPDFExporter.allow_chromium_download=True]
--disable-chromium-sandbox
    Disable chromium security sandbox when converting to PDF..
    Equivalent to: [--WebPDFExporter.disable_sandbox=True]
--show-input
    Shows code input. This flag is only useful for dejavu users.
    Equivalent to: [--TemplateExporter.exclude_input=False]
--embed-images
    Embed the images as base64 dataurls in the output. This flag is only useful for the HTML/WebPDF/Slides exports.
    Equivalent to: [--HTMLExporter.embed_images=True]
--sanitize-html
    Whether the HTML in Markdown cells and cell outputs should be sanitized..
    Equivalent to: [--HTMLExporter.sanitize_html=True]
--log-level=<Enum>
    Set the log level by value or name.
    Choices: any of [0, 10, 20, 30, 40, 50, 'DEBUG', 'INFO', 'WARN', 'ERROR', 'CRITICAL']
    Default: 30
    Equivalent to: [--Application.log_level]
--config=<Unicode>
    Full path of a config file.
    Default: ''
    Equivalent to: [--JupyterApp.config_file]
--to=<Unicode>
    The export format to be used, either one of the built-in formats
            ['asciidoc', 'custom', 'html', 'latex', 'markdown', 'notebook', 'pdf', 'python', 'rst', 'script', 'slides', 'webpdf']
            or a dotted object name that represents the import path for an
            ``Exporter`` class
    Default: ''
    Equivalent to: [--NbConvertApp.export_format]
--template=<Unicode>
    Name of the template to use
    Default: ''
    Equivalent to: [--TemplateExporter.template_name]
--template-file=<Unicode>
    Name of the template file to use
    Default: None
    Equivalent to: [--TemplateExporter.template_file]
--theme=<Unicode>
    Template specific theme(e.g. the name of a JupyterLab CSS theme distributed
    as prebuilt extension for the lab template)
    Default: 'light'
    Equivalent to: [--HTMLExporter.theme]
--sanitize_html=<Bool>
    Whether the HTML in Markdown cells and cell outputs should be sanitized.This
    should be set to True by nbviewer or similar tools.
    Default: False
    Equivalent to: [--HTMLExporter.sanitize_html]
--writer=<DottedObjectName>
    Writer class used to write the
                                        results of the conversion
    Default: 'FilesWriter'
    Equivalent to: [--NbConvertApp.writer_class]
--post=<DottedOrNone>
    PostProcessor class used to write the
                                        results of the conversion
    Default: ''
    Equivalent to: [--NbConvertApp.postprocessor_class]
--output=<Unicode>
    overwrite base name use for output files.
                can only be used when converting one notebook at a time.
    Default: ''
    Equivalent to: [--NbConvertApp.output_base]
--output-dir=<Unicode>
    Directory to write output(s) to. Defaults
                                  to output to the directory of each notebook. To recover
                                  previous default behaviour (outputting to the current
                                  working directory) use . as the flag value.
    Default: ''
    Equivalent to: [--FilesWriter.build_directory]
--reveal-prefix=<Unicode>
    The URL prefix for reveal.js (version 3.x).
            This defaults to the reveal CDN, but can be any url pointing to a copy
            of reveal.js.
            For speaker notes to work, this must be a relative path to a local
            copy of reveal.js: e.g., "reveal.js".
            If a relative path is given, it must be a subdirectory of the
            current directory (from which the server is run).
            See the usage documentation
            (https://nbconvert.readthedocs.io/en/latest/usage.html#reveal-js-html-slideshow)
            for more details.
    Default: ''
    Equivalent to: [--SlidesExporter.reveal_url_prefix]
--nbformat=<Enum>
    The nbformat version to write.
            Use this to downgrade notebooks.
    Choices: any of [1, 2, 3, 4]
    Default: 4
    Equivalent to: [--NotebookExporter.nbformat_version]

Examples
--------

    The simplest way to use nbconvert is

            > jupyter nbconvert mynotebook.ipynb --to html

            Options include ['asciidoc', 'custom', 'html', 'latex', 'markdown', 'notebook', 'pdf', 'python', 'rst', 'script', 'slides', 'webpdf'].

            > jupyter nbconvert --to latex mynotebook.ipynb

            Both HTML and LaTeX support multiple output templates. LaTeX includes
            'base', 'article' and 'report'.  HTML includes 'basic', 'lab' and
            'classic'. You can specify the flavor of the format used.

            > jupyter nbconvert --to html --template lab mynotebook.ipynb

            You can also pipe the output to stdout, rather than a file

            > jupyter nbconvert mynotebook.ipynb --stdout

            PDF is generated via latex

            > jupyter nbconvert mynotebook.ipynb --to pdf

            You can get (and serve) a Reveal.js-powered slideshow

            > jupyter nbconvert myslides.ipynb --to slides --post serve

            Multiple notebooks can be given at the command line in a couple of
            different ways:

            > jupyter nbconvert notebook*.ipynb
            > jupyter nbconvert notebook1.ipynb notebook2.ipynb

            or you can specify the notebooks list in a config file, containing::

                c.NbConvertApp.notebooks = ["my_notebook.ipynb"]

            > jupyter nbconvert --config mycfg.py

To see all available configurables, use `--help-all`.